Merge 75f00f756f into db7b106f1d

Merge pull request #10611 from liamwhite/audio-deadlock
audio_renderer: resolve adsp thread deadlock shutdown
2023-06-05 21:22:20 -04:00 · 2023-06-05 17:15:19 -07:00 · 2023-06-04 13:24:47 -07:00 · 2023-06-04 13:00:10 -04:00 · 2023-06-03 23:23:45 -07:00 · 2023-06-03 23:22:24 -07:00
164 changed files with 21943 additions and 16671 deletions
@@ -26,7 +26,11 @@ $env:BUILD_ZIP = $MSVC_BUILD_ZIP
 $env:BUILD_SYMBOLS = $MSVC_BUILD_PDB
 $env:BUILD_UPDATE = $MSVC_SEVENZIP

-$BUILD_DIR = ".\build\bin\Release"
+if (Test-Path -Path ".\build\bin\Release") {
+    $BUILD_DIR = ".\build\bin\Release"
+} else {
+    $BUILD_DIR = ".\build\bin\"
+}

 # Cleanup unneeded data in submodules
 git submodule foreach git clean -fxd
@@ -43,7 +43,7 @@ body:
    id: log
    attributes:
      label: Log File
-      description: A log file will help our developers to better diagnose and fix the issue.
+      description: A log file will help our developers to better diagnose and fix the issue. Instructions can be found [here](https://yuzu-emu.org/help/reference/log-files).
    validations:
      required: true
  - type: textarea
@@ -453,6 +453,7 @@ endif()
 # List of all FFmpeg components required
 set(FFmpeg_COMPONENTS
    avcodec
+    avfilter
    avutil
    swscale)

@@ -3,8 +3,8 @@

 function(copy_yuzu_FFmpeg_deps target_dir)
    include(WindowsCopyFiles)
-    set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
+    set(DLL_DEST "$<TARGET_FILE_DIR:${target_dir}>/")
    file(READ "${FFmpeg_PATH}/requirements.txt" FFmpeg_REQUIRED_DLLS)
    string(STRIP "${FFmpeg_REQUIRED_DLLS}" FFmpeg_REQUIRED_DLLS)
-    windows_copy_files(${target_dir} ${FFmpeg_DLL_DIR} ${DLL_DEST} ${FFmpeg_REQUIRED_DLLS})
+    windows_copy_files(${target_dir} ${FFmpeg_LIBRARY_DIR} ${DLL_DEST} ${FFmpeg_REQUIRED_DLLS})
 endfunction(copy_yuzu_FFmpeg_deps)
@@ -4,7 +4,7 @@
 function(copy_yuzu_Qt5_deps target_dir)
    include(WindowsCopyFiles)
    if (MSVC)
-        set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
+        set(DLL_DEST "$<TARGET_FILE_DIR:${target_dir}>/")
        set(Qt5_DLL_DIR "${Qt5_DIR}/../../../bin")
    else()
        set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/")
@@ -3,6 +3,6 @@

 function(copy_yuzu_SDL_deps target_dir)
    include(WindowsCopyFiles)
-    set(DLL_DEST "${CMAKE_BINARY_DIR}/bin/$<CONFIG>/")
+    set(DLL_DEST "$<TARGET_FILE_DIR:${target_dir}>/")
    windows_copy_files(${target_dir} ${SDL2_DLL_DIR} ${DLL_DEST} SDL2.dll)
 endfunction(copy_yuzu_SDL_deps)
@@ -13,7 +13,7 @@ SPDX-License-Identifier: GPL-2.0-or-later

 <h4 align="center"><b>yuzu</b> is the world's most popular, open-source, Nintendo Switch emulator — started by the creators of <a href="https://citra-emu.org" target="_blank">Citra</a>.
 <br>
-It is written in C++ with portability in mind, and we actively maintain builds for Windows and Linux.
+It is written in C++ with portability in mind, and we actively maintain builds for Windows, Linux and Android.
 </h4>

 <p align="center">
@@ -83,5 +83,3 @@ If you wish to support us a different way, please join our [Discord](https://dis
 ## License

 yuzu is licensed under the GPLv3 (or any later version). Refer to the [LICENSE.txt](https://github.com/yuzu-emu/yuzu/blob/master/LICENSE.txt) file.
-
-The [Skyline-Emulator Team](https://github.com/skyline-emu/skyline) may choose to use the code from these contributors under the GPL-3.0-or-later OR MPL-2.0: [FernandoS27](https://github.com/FernandoS27), [lioncash](https://github.com/lioncash), [bunnei](https://github.com/bunnei), [ReinUsesLisp](https://github.com/ReinUsesLisp), [Morph1984](https://github.com/Morph1984), [ogniK5377](https://github.com/ogniK5377), [german77](https://github.com/german77), [ameerj](https://github.com/ameerj), [Kelebek1](https://github.com/Kelebek1) and [lat9nq](https://github.com/lat9nq)
@@ -8,15 +8,21 @@ set(CMAKE_POLICY_DEFAULT_CMP0077 NEW)
 # Disable tests in all externals supporting the standard option name
 set(BUILD_TESTING OFF)

+# Build only static externals
+set(BUILD_SHARED_LIBS OFF)
+
+# Skip install rules for all externals
+set_directory_properties(PROPERTIES EXCLUDE_FROM_ALL ON)
+
 # xbyak
 if ((ARCHITECTURE_x86 OR ARCHITECTURE_x86_64) AND NOT TARGET xbyak::xbyak)
-    add_subdirectory(xbyak EXCLUDE_FROM_ALL)
+    add_subdirectory(xbyak)
 endif()

 # Dynarmic
 if ((ARCHITECTURE_x86_64 OR ARCHITECTURE_arm64) AND NOT TARGET dynarmic::dynarmic)
    set(DYNARMIC_IGNORE_ASSERTS ON)
-    add_subdirectory(dynarmic EXCLUDE_FROM_ALL)
+    add_subdirectory(dynarmic)
    add_library(dynarmic::dynarmic ALIAS dynarmic)
 endif()

@@ -34,7 +40,7 @@ if (NOT TARGET inih::INIReader)
 endif()

 # mbedtls
-add_subdirectory(mbedtls EXCLUDE_FROM_ALL)
+add_subdirectory(mbedtls)
 target_include_directories(mbedtls PUBLIC ./mbedtls/include)

 # MicroProfile
@@ -48,7 +54,7 @@ endif()

 # libusb
 if (ENABLE_LIBUSB AND NOT TARGET libusb::usb)
-    add_subdirectory(libusb EXCLUDE_FROM_ALL)
+    add_subdirectory(libusb)
 endif()

 # SDL2
@@ -67,18 +73,16 @@ if (YUZU_USE_EXTERNAL_SDL2)

        set(HIDAPI ON)
    endif()
-    set(SDL_STATIC ON)
-    set(SDL_SHARED OFF)
    if (APPLE)
        set(SDL_FILE ON)
    endif()

-    add_subdirectory(SDL EXCLUDE_FROM_ALL)
+    add_subdirectory(SDL)
 endif()

 # ENet
 if (NOT TARGET enet::enet)
-    add_subdirectory(enet EXCLUDE_FROM_ALL)
+    add_subdirectory(enet)
    target_include_directories(enet INTERFACE ./enet/include)
    add_library(enet::enet ALIAS enet)
 endif()
@@ -86,24 +90,26 @@ endif()
 # Cubeb
 if (ENABLE_CUBEB AND NOT TARGET cubeb::cubeb)
    set(BUILD_TESTS OFF)
-    add_subdirectory(cubeb EXCLUDE_FROM_ALL)
+    set(BUILD_TOOLS OFF)
+    add_subdirectory(cubeb)
    add_library(cubeb::cubeb ALIAS cubeb)
 endif()

 # DiscordRPC
 if (USE_DISCORD_PRESENCE AND NOT TARGET DiscordRPC::discord-rpc)
-    add_subdirectory(discord-rpc EXCLUDE_FROM_ALL)
+    set(BUILD_EXAMPLES OFF)
+    add_subdirectory(discord-rpc)
    target_include_directories(discord-rpc INTERFACE ./discord-rpc/include)
    add_library(DiscordRPC::discord-rpc ALIAS discord-rpc)
 endif()

 # Sirit
-add_subdirectory(sirit EXCLUDE_FROM_ALL)
+add_subdirectory(sirit)

 # httplib
 if (ENABLE_WEB_SERVICE AND NOT TARGET httplib::httplib)
    set(HTTPLIB_REQUIRE_OPENSSL ON)
-    add_subdirectory(cpp-httplib EXCLUDE_FROM_ALL)
+    add_subdirectory(cpp-httplib)
 endif()

 # cpp-jwt
@@ -111,12 +117,12 @@ if (ENABLE_WEB_SERVICE AND NOT TARGET cpp-jwt::cpp-jwt)
    set(CPP_JWT_BUILD_EXAMPLES OFF)
    set(CPP_JWT_BUILD_TESTS OFF)
    set(CPP_JWT_USE_VENDORED_NLOHMANN_JSON OFF)
-    add_subdirectory(cpp-jwt EXCLUDE_FROM_ALL)
+    add_subdirectory(cpp-jwt)
 endif()

 # Opus
 if (NOT TARGET Opus::opus)
-    add_subdirectory(opus EXCLUDE_FROM_ALL)
+    add_subdirectory(opus)
 endif()

 # FFMpeg
@@ -130,12 +136,14 @@ endif()

 # Vulkan-Headers
 if (YUZU_USE_EXTERNAL_VULKAN_HEADERS)
-    add_subdirectory(Vulkan-Headers EXCLUDE_FROM_ALL)
+    add_subdirectory(Vulkan-Headers)
 endif()

 if (NOT TARGET LLVM::Demangle)
-    add_library(demangle STATIC)
+    add_library(demangle demangle/ItaniumDemangle.cpp)
    target_include_directories(demangle PUBLIC ./demangle)
-    target_sources(demangle PRIVATE demangle/ItaniumDemangle.cpp)
    add_library(LLVM::Demangle ALIAS demangle)
 endif()
+
+add_library(stb stb/stb_dxt.cpp)
+target_include_directories(stb PUBLIC ./stb)
@@ -131,7 +131,6 @@ if (NOT WIN32)
        COMMAND
            /bin/bash ${FFmpeg_PREFIX}/configure
                --disable-avdevice
-                --disable-avfilter
                --disable-avformat
                --disable-doc
                --disable-everything
@@ -143,6 +142,7 @@ if (NOT WIN32)
                --enable-decoder=h264
                --enable-decoder=vp8
                --enable-decoder=vp9
+                --enable-filter=yadif
                --cc="${FFmpeg_CC}"
                --cxx="${FFmpeg_CXX}"
                ${FFmpeg_HWACCEL_FLAGS}
@@ -199,7 +199,7 @@ if (NOT WIN32)
    endif()
 else(WIN32)
    # Use yuzu FFmpeg binaries
-    set(FFmpeg_EXT_NAME "ffmpeg-4.4")
+    set(FFmpeg_EXT_NAME "ffmpeg-5.1.3")
    set(FFmpeg_PATH "${CMAKE_BINARY_DIR}/externals/${FFmpeg_EXT_NAME}")
    download_bundled_external("ffmpeg/" ${FFmpeg_EXT_NAME} "")
    set(FFmpeg_FOUND YES)
@@ -210,6 +210,7 @@ else(WIN32)
    set(FFmpeg_LIBRARIES
    ${FFmpeg_LIBRARY_DIR}/swscale.lib
    ${FFmpeg_LIBRARY_DIR}/avcodec.lib
+    ${FFmpeg_LIBRARY_DIR}/avfilter.lib
    ${FFmpeg_LIBRARY_DIR}/avutil.lib
    CACHE PATH "Paths to FFmpeg libraries" FORCE)
    # exported variables
@@ -1,7 +1,7 @@
 # SPDX-FileCopyrightText: 2015 Yuri Kunde Schlesner <yuriks@yuriks.net>
 # SPDX-License-Identifier: GPL-2.0-or-later

-add_library(glad STATIC
+add_library(glad
    src/glad.c
    include/KHR/khrplatform.h
    include/glad/glad.h
@@ -122,7 +122,7 @@ else() # MINGW OR (${CMAKE_SYSTEM_NAME} MATCHES "Linux")
        add_compile_options(/utf-8)
    endif()

-    add_library(usb STATIC EXCLUDE_FROM_ALL
+    add_library(usb
        libusb/libusb/core.c
        libusb/libusb/core.c
        libusb/libusb/descriptor.c
@@ -1697,7 +1697,13 @@ void MicroProfileFlip()
                            {
                                int nTimer = MicroProfileLogTimerIndex(LE);
                                uint8_t nGroup = pTimerToGroup[nTimer];
-                                MP_ASSERT(nStackPos < MICROPROFILE_STACK_MAX);
+
+                                // To avoid crashing due to OOB memory accesses/asserts
+                                // simply skip this iteration
+                                // MP_ASSERT(nStackPos < MICROPROFILE_STACK_MAX);
+                                if (nStackPos >= MICROPROFILE_STACK_MAX) {
+                                    break;
+                                }
                                MP_ASSERT(nGroup < MICROPROFILE_MAX_GROUPS);
                                pGroupStackPos[nGroup]++;
                                pStack[nStackPos++] = k;
@@ -23,7 +23,7 @@ else()
    endif()
 endif()

-add_library(opus STATIC
+add_library(opus
    # CELT sources
    opus/celt/bands.c
    opus/celt/celt.c
@@ -0,0 +1,765 @@
+// SPDX-FileCopyrightText: fabian "ryg" giesen
+// SPDX-License-Identifier: MIT
+
+// stb_dxt.h - v1.12 - DXT1/DXT5 compressor
+
+#include <stb_dxt.h>
+
+#include <stdlib.h>
+#include <string.h>
+
+#if !defined(STBD_FABS)
+#include <math.h>
+#endif
+
+#ifndef STBD_FABS
+#define STBD_FABS(x) fabs(x)
+#endif
+
+static const unsigned char stb__OMatch5[256][2] = {
+    {0, 0},   {0, 0},   {0, 1},   {0, 1},   {1, 0},   {1, 0},   {1, 0},   {1, 1},   {1, 1},
+    {1, 1},   {1, 2},   {0, 4},   {2, 1},   {2, 1},   {2, 1},   {2, 2},   {2, 2},   {2, 2},
+    {2, 3},   {1, 5},   {3, 2},   {3, 2},   {4, 0},   {3, 3},   {3, 3},   {3, 3},   {3, 4},
+    {3, 4},   {3, 4},   {3, 5},   {4, 3},   {4, 3},   {5, 2},   {4, 4},   {4, 4},   {4, 5},
+    {4, 5},   {5, 4},   {5, 4},   {5, 4},   {6, 3},   {5, 5},   {5, 5},   {5, 6},   {4, 8},
+    {6, 5},   {6, 5},   {6, 5},   {6, 6},   {6, 6},   {6, 6},   {6, 7},   {5, 9},   {7, 6},
+    {7, 6},   {8, 4},   {7, 7},   {7, 7},   {7, 7},   {7, 8},   {7, 8},   {7, 8},   {7, 9},
+    {8, 7},   {8, 7},   {9, 6},   {8, 8},   {8, 8},   {8, 9},   {8, 9},   {9, 8},   {9, 8},
+    {9, 8},   {10, 7},  {9, 9},   {9, 9},   {9, 10},  {8, 12},  {10, 9},  {10, 9},  {10, 9},
+    {10, 10}, {10, 10}, {10, 10}, {10, 11}, {9, 13},  {11, 10}, {11, 10}, {12, 8},  {11, 11},
+    {11, 11}, {11, 11}, {11, 12}, {11, 12}, {11, 12}, {11, 13}, {12, 11}, {12, 11}, {13, 10},
+    {12, 12}, {12, 12}, {12, 13}, {12, 13}, {13, 12}, {13, 12}, {13, 12}, {14, 11}, {13, 13},
+    {13, 13}, {13, 14}, {12, 16}, {14, 13}, {14, 13}, {14, 13}, {14, 14}, {14, 14}, {14, 14},
+    {14, 15}, {13, 17}, {15, 14}, {15, 14}, {16, 12}, {15, 15}, {15, 15}, {15, 15}, {15, 16},
+    {15, 16}, {15, 16}, {15, 17}, {16, 15}, {16, 15}, {17, 14}, {16, 16}, {16, 16}, {16, 17},
+    {16, 17}, {17, 16}, {17, 16}, {17, 16}, {18, 15}, {17, 17}, {17, 17}, {17, 18}, {16, 20},
+    {18, 17}, {18, 17}, {18, 17}, {18, 18}, {18, 18}, {18, 18}, {18, 19}, {17, 21}, {19, 18},
+    {19, 18}, {20, 16}, {19, 19}, {19, 19}, {19, 19}, {19, 20}, {19, 20}, {19, 20}, {19, 21},
+    {20, 19}, {20, 19}, {21, 18}, {20, 20}, {20, 20}, {20, 21}, {20, 21}, {21, 20}, {21, 20},
+    {21, 20}, {22, 19}, {21, 21}, {21, 21}, {21, 22}, {20, 24}, {22, 21}, {22, 21}, {22, 21},
+    {22, 22}, {22, 22}, {22, 22}, {22, 23}, {21, 25}, {23, 22}, {23, 22}, {24, 20}, {23, 23},
+    {23, 23}, {23, 23}, {23, 24}, {23, 24}, {23, 24}, {23, 25}, {24, 23}, {24, 23}, {25, 22},
+    {24, 24}, {24, 24}, {24, 25}, {24, 25}, {25, 24}, {25, 24}, {25, 24}, {26, 23}, {25, 25},
+    {25, 25}, {25, 26}, {24, 28}, {26, 25}, {26, 25}, {26, 25}, {26, 26}, {26, 26}, {26, 26},
+    {26, 27}, {25, 29}, {27, 26}, {27, 26}, {28, 24}, {27, 27}, {27, 27}, {27, 27}, {27, 28},
+    {27, 28}, {27, 28}, {27, 29}, {28, 27}, {28, 27}, {29, 26}, {28, 28}, {28, 28}, {28, 29},
+    {28, 29}, {29, 28}, {29, 28}, {29, 28}, {30, 27}, {29, 29}, {29, 29}, {29, 30}, {29, 30},
+    {30, 29}, {30, 29}, {30, 29}, {30, 30}, {30, 30}, {30, 30}, {30, 31}, {30, 31}, {31, 30},
+    {31, 30}, {31, 30}, {31, 31}, {31, 31},
+};
+static const unsigned char stb__OMatch6[256][2] = {
+    {0, 0},   {0, 1},   {1, 0},   {1, 1},   {1, 1},   {1, 2},   {2, 1},   {2, 2},   {2, 2},
+    {2, 3},   {3, 2},   {3, 3},   {3, 3},   {3, 4},   {4, 3},   {4, 4},   {4, 4},   {4, 5},
+    {5, 4},   {5, 5},   {5, 5},   {5, 6},   {6, 5},   {6, 6},   {6, 6},   {6, 7},   {7, 6},
+    {7, 7},   {7, 7},   {7, 8},   {8, 7},   {8, 8},   {8, 8},   {8, 9},   {9, 8},   {9, 9},
+    {9, 9},   {9, 10},  {10, 9},  {10, 10}, {10, 10}, {10, 11}, {11, 10}, {8, 16},  {11, 11},
+    {11, 12}, {12, 11}, {9, 17},  {12, 12}, {12, 13}, {13, 12}, {11, 16}, {13, 13}, {13, 14},
+    {14, 13}, {12, 17}, {14, 14}, {14, 15}, {15, 14}, {14, 16}, {15, 15}, {15, 16}, {16, 14},
+    {16, 15}, {17, 14}, {16, 16}, {16, 17}, {17, 16}, {18, 15}, {17, 17}, {17, 18}, {18, 17},
+    {20, 14}, {18, 18}, {18, 19}, {19, 18}, {21, 15}, {19, 19}, {19, 20}, {20, 19}, {20, 20},
+    {20, 20}, {20, 21}, {21, 20}, {21, 21}, {21, 21}, {21, 22}, {22, 21}, {22, 22}, {22, 22},
+    {22, 23}, {23, 22}, {23, 23}, {23, 23}, {23, 24}, {24, 23}, {24, 24}, {24, 24}, {24, 25},
+    {25, 24}, {25, 25}, {25, 25}, {25, 26}, {26, 25}, {26, 26}, {26, 26}, {26, 27}, {27, 26},
+    {24, 32}, {27, 27}, {27, 28}, {28, 27}, {25, 33}, {28, 28}, {28, 29}, {29, 28}, {27, 32},
+    {29, 29}, {29, 30}, {30, 29}, {28, 33}, {30, 30}, {30, 31}, {31, 30}, {30, 32}, {31, 31},
+    {31, 32}, {32, 30}, {32, 31}, {33, 30}, {32, 32}, {32, 33}, {33, 32}, {34, 31}, {33, 33},
+    {33, 34}, {34, 33}, {36, 30}, {34, 34}, {34, 35}, {35, 34}, {37, 31}, {35, 35}, {35, 36},
+    {36, 35}, {36, 36}, {36, 36}, {36, 37}, {37, 36}, {37, 37}, {37, 37}, {37, 38}, {38, 37},
+    {38, 38}, {38, 38}, {38, 39}, {39, 38}, {39, 39}, {39, 39}, {39, 40}, {40, 39}, {40, 40},
+    {40, 40}, {40, 41}, {41, 40}, {41, 41}, {41, 41}, {41, 42}, {42, 41}, {42, 42}, {42, 42},
+    {42, 43}, {43, 42}, {40, 48}, {43, 43}, {43, 44}, {44, 43}, {41, 49}, {44, 44}, {44, 45},
+    {45, 44}, {43, 48}, {45, 45}, {45, 46}, {46, 45}, {44, 49}, {46, 46}, {46, 47}, {47, 46},
+    {46, 48}, {47, 47}, {47, 48}, {48, 46}, {48, 47}, {49, 46}, {48, 48}, {48, 49}, {49, 48},
+    {50, 47}, {49, 49}, {49, 50}, {50, 49}, {52, 46}, {50, 50}, {50, 51}, {51, 50}, {53, 47},
+    {51, 51}, {51, 52}, {52, 51}, {52, 52}, {52, 52}, {52, 53}, {53, 52}, {53, 53}, {53, 53},
+    {53, 54}, {54, 53}, {54, 54}, {54, 54}, {54, 55}, {55, 54}, {55, 55}, {55, 55}, {55, 56},
+    {56, 55}, {56, 56}, {56, 56}, {56, 57}, {57, 56}, {57, 57}, {57, 57}, {57, 58}, {58, 57},
+    {58, 58}, {58, 58}, {58, 59}, {59, 58}, {59, 59}, {59, 59}, {59, 60}, {60, 59}, {60, 60},
+    {60, 60}, {60, 61}, {61, 60}, {61, 61}, {61, 61}, {61, 62}, {62, 61}, {62, 62}, {62, 62},
+    {62, 63}, {63, 62}, {63, 63}, {63, 63},
+};
+
+static int stb__Mul8Bit(int a, int b) {
+    int t = a * b + 128;
+    return (t + (t >> 8)) >> 8;
+}
+
+static void stb__From16Bit(unsigned char* out, unsigned short v) {
+    int rv = (v & 0xf800) >> 11;
+    int gv = (v & 0x07e0) >> 5;
+    int bv = (v & 0x001f) >> 0;
+
+    // expand to 8 bits via bit replication
+    out[0] = static_cast<unsigned char>((rv * 33) >> 2);
+    out[1] = static_cast<unsigned char>((gv * 65) >> 4);
+    out[2] = static_cast<unsigned char>((bv * 33) >> 2);
+    out[3] = 0;
+}
+
+static unsigned short stb__As16Bit(int r, int g, int b) {
+    return (unsigned short)((stb__Mul8Bit(r, 31) << 11) + (stb__Mul8Bit(g, 63) << 5) +
+                            stb__Mul8Bit(b, 31));
+}
+
+// linear interpolation at 1/3 point between a and b, using desired rounding
+// type
+static int stb__Lerp13(int a, int b) {
+#ifdef STB_DXT_USE_ROUNDING_BIAS
+    // with rounding bias
+    return a + stb__Mul8Bit(b - a, 0x55);
+#else
+    // without rounding bias
+    // replace "/ 3" by "* 0xaaab) >> 17" if your compiler sucks or you really
+    // need every ounce of speed.
+    return (2 * a + b) / 3;
+#endif
+}
+
+// linear interpolation at 1/2 point between a and b
+static int stb__Lerp12(int a, int b) {
+    return (a + b) / 2;
+}
+
+// lerp RGB color
+static void stb__Lerp13RGB(unsigned char* out, unsigned char* p1, unsigned char* p2) {
+    out[0] = (unsigned char)stb__Lerp13(p1[0], p2[0]);
+    out[1] = (unsigned char)stb__Lerp13(p1[1], p2[1]);
+    out[2] = (unsigned char)stb__Lerp13(p1[2], p2[2]);
+}
+
+static void stb__Lerp12RGB(unsigned char* out, unsigned char* p1, unsigned char* p2) {
+    out[0] = (unsigned char)stb__Lerp12(p1[0], p2[0]);
+    out[1] = (unsigned char)stb__Lerp12(p1[1], p2[1]);
+    out[2] = (unsigned char)stb__Lerp12(p1[2], p2[2]);
+}
+
+/****************************************************************************/
+
+static void stb__Eval4Colors(unsigned char* color, unsigned short c0, unsigned short c1) {
+    stb__From16Bit(color + 0, c0);
+    stb__From16Bit(color + 4, c1);
+    stb__Lerp13RGB(color + 8, color + 0, color + 4);
+    stb__Lerp13RGB(color + 12, color + 4, color + 0);
+}
+
+static void stb__Eval3Colors(unsigned char* color, unsigned short c0, unsigned short c1) {
+    stb__From16Bit(color + 0, c0);
+    stb__From16Bit(color + 4, c1);
+    stb__Lerp12RGB(color + 8, color + 0, color + 4);
+}
+
+// The color matching function
+static unsigned int stb__MatchColorsBlock(unsigned char* block, unsigned char* color) {
+    unsigned int mask = 0;
+    int dirr = color[0 * 4 + 0] - color[1 * 4 + 0];
+    int dirg = color[0 * 4 + 1] - color[1 * 4 + 1];
+    int dirb = color[0 * 4 + 2] - color[1 * 4 + 2];
+    int dots[16];
+    int stops[4];
+    int i;
+    int c0Point, halfPoint, c3Point;
+
+    for (i = 0; i < 16; i++)
+        dots[i] = block[i * 4 + 0] * dirr + block[i * 4 + 1] * dirg + block[i * 4 + 2] * dirb;
+
+    for (i = 0; i < 4; i++)
+        stops[i] = color[i * 4 + 0] * dirr + color[i * 4 + 1] * dirg + color[i * 4 + 2] * dirb;
+
+    // think of the colors as arranged on a line; project point onto that line,
+    // then choose next color out of available ones. we compute the crossover
+    // points for "best color in top half"/"best in bottom half" and then the same
+    // inside that subinterval.
+    //
+    // relying on this 1d approximation isn't always optimal in terms of euclidean
+    // distance, but it's very close and a lot faster.
+    // http://cbloomrants.blogspot.com/2008/12/12-08-08-dxtc-summary.html
+
+    c0Point = (stops[1] + stops[3]);
+    halfPoint = (stops[3] + stops[2]);
+    c3Point = (stops[2] + stops[0]);
+
+    for (i = 15; i >= 0; i--) {
+        int dot = dots[i] * 2;
+        mask <<= 2;
+
+        if (dot < halfPoint)
+            mask |= (dot < c0Point) ? 1 : 3;
+        else
+            mask |= (dot < c3Point) ? 2 : 0;
+    }
+
+    return mask;
+}
+
+static unsigned int stb__MatchColorsAlphaBlock(unsigned char* block, unsigned char* color) {
+    unsigned int mask = 0;
+    int dirr = color[0 * 4 + 0] - color[1 * 4 + 0];
+    int dirg = color[0 * 4 + 1] - color[1 * 4 + 1];
+    int dirb = color[0 * 4 + 2] - color[1 * 4 + 2];
+    int dots[16];
+    int stops[3];
+    int i;
+    int c0Point, c2Point;
+
+    for (i = 0; i < 16; i++)
+        dots[i] = block[i * 4 + 0] * dirr + block[i * 4 + 1] * dirg + block[i * 4 + 2] * dirb;
+
+    for (i = 0; i < 3; i++)
+        stops[i] = color[i * 4 + 0] * dirr + color[i * 4 + 1] * dirg + color[i * 4 + 2] * dirb;
+
+    c0Point = (stops[1] + stops[2]);
+    c2Point = (stops[2] + stops[0]);
+
+    for (i = 15; i >= 0; i--) {
+        int dot = dots[i] * 2;
+        mask <<= 2;
+
+        if (block[i * 4 + 3] == 0)
+            mask |= 3;
+        else if (dot < c2Point)
+            mask |= (dot < c0Point) ? 0 : 2;
+        else
+            mask |= (dot < c0Point) ? 1 : 0;
+    }
+
+    return mask;
+}
+
+static void stb__ReorderColors(unsigned short* pmax16, unsigned short* pmin16) {
+    if (*pmin16 < *pmax16) {
+        unsigned short t = *pmin16;
+        *pmin16 = *pmax16;
+        *pmax16 = t;
+    }
+}
+
+static void stb__FinalizeColors(unsigned short* pmax16, unsigned short* pmin16,
+                                unsigned int* pmask) {
+    if (*pmax16 < *pmin16) {
+        unsigned short t = *pmin16;
+        *pmin16 = *pmax16;
+        *pmax16 = t;
+        *pmask ^= 0x55555555;
+    }
+}
+
+// The color optimization function. (Clever code, part 1)
+static void stb__OptimizeColorsBlock(unsigned char* block, unsigned short* pmax16,
+                                     unsigned short* pmin16) {
+    int mind, maxd;
+    unsigned char *minp, *maxp;
+    double magn;
+    int v_r, v_g, v_b;
+    static const int nIterPower = 4;
+    float covf[6], vfr, vfg, vfb;
+
+    // determine color distribution
+    int cov[6];
+    int mu[3], min[3], max[3];
+    int ch, i, iter;
+
+    for (ch = 0; ch < 3; ch++) {
+        const unsigned char* bp = ((const unsigned char*)block) + ch;
+        int muv, minv, maxv;
+
+        muv = minv = maxv = bp[0];
+        for (i = 4; i < 64; i += 4) {
+            muv += bp[i];
+            if (bp[i] < minv)
+                minv = bp[i];
+            else if (bp[i] > maxv)
+                maxv = bp[i];
+        }
+
+        mu[ch] = (muv + 8) >> 4;
+        min[ch] = minv;
+        max[ch] = maxv;
+    }
+
+    // determine covariance matrix
+    for (i = 0; i < 6; i++)
+        cov[i] = 0;
+
+    for (i = 0; i < 16; i++) {
+        int r = block[i * 4 + 0] - mu[0];
+        int g = block[i * 4 + 1] - mu[1];
+        int b = block[i * 4 + 2] - mu[2];
+
+        cov[0] += r * r;
+        cov[1] += r * g;
+        cov[2] += r * b;
+        cov[3] += g * g;
+        cov[4] += g * b;
+        cov[5] += b * b;
+    }
+
+    // convert covariance matrix to float, find principal axis via power iter
+    for (i = 0; i < 6; i++)
+        covf[i] = static_cast<float>(cov[i]) / 255.0f;
+
+    vfr = (float)(max[0] - min[0]);
+    vfg = (float)(max[1] - min[1]);
+    vfb = (float)(max[2] - min[2]);
+
+    for (iter = 0; iter < nIterPower; iter++) {
+        float r = vfr * covf[0] + vfg * covf[1] + vfb * covf[2];
+        float g = vfr * covf[1] + vfg * covf[3] + vfb * covf[4];
+        float b = vfr * covf[2] + vfg * covf[4] + vfb * covf[5];
+
+        vfr = r;
+        vfg = g;
+        vfb = b;
+    }
+
+    magn = STBD_FABS(vfr);
+    if (STBD_FABS(vfg) > magn)
+        magn = STBD_FABS(vfg);
+    if (STBD_FABS(vfb) > magn)
+        magn = STBD_FABS(vfb);
+
+    if (magn < 4.0f) { // too small, default to luminance
+        v_r = 299;     // JPEG YCbCr luma coefs, scaled by 1000.
+        v_g = 587;
+        v_b = 114;
+    } else {
+        magn = 512.0 / magn;
+        v_r = (int)(vfr * magn);
+        v_g = (int)(vfg * magn);
+        v_b = (int)(vfb * magn);
+    }
+
+    minp = maxp = block;
+    mind = maxd = block[0] * v_r + block[1] * v_g + block[2] * v_b;
+    // Pick colors at extreme points
+    for (i = 1; i < 16; i++) {
+        int dot = block[i * 4 + 0] * v_r + block[i * 4 + 1] * v_g + block[i * 4 + 2] * v_b;
+
+        if (dot < mind) {
+            mind = dot;
+            minp = block + i * 4;
+        }
+
+        if (dot > maxd) {
+            maxd = dot;
+            maxp = block + i * 4;
+        }
+    }
+
+    *pmax16 = stb__As16Bit(maxp[0], maxp[1], maxp[2]);
+    *pmin16 = stb__As16Bit(minp[0], minp[1], minp[2]);
+    stb__ReorderColors(pmax16, pmin16);
+}
+
+static void stb__OptimizeColorsAlphaBlock(unsigned char* block, unsigned short* pmax16,
+                                          unsigned short* pmin16) {
+    int mind, maxd;
+    unsigned char *minp, *maxp;
+    double magn;
+    int v_r, v_g, v_b;
+    static const int nIterPower = 4;
+    float covf[6], vfr, vfg, vfb;
+
+    // determine color distribution
+    int cov[6];
+    int mu[3], min[3], max[3];
+    int ch, i, iter;
+
+    for (ch = 0; ch < 3; ch++) {
+        const unsigned char* bp = ((const unsigned char*)block) + ch;
+        int muv = 0, minv = 256, maxv = -1;
+        int num = 0;
+
+        for (i = 0; i < 64; i += 4) {
+            if (bp[3 - ch] == 0) {
+                continue;
+            }
+
+            muv += bp[i];
+            if (bp[i] < minv)
+                minv = bp[i];
+            else if (bp[i] > maxv)
+                maxv = bp[i];
+
+            num++;
+        }
+
+        mu[ch] = num > 0 ? (muv + 8) / num : 0;
+        min[ch] = minv;
+        max[ch] = maxv;
+    }
+
+    // determine covariance matrix
+    for (i = 0; i < 6; i++)
+        cov[i] = 0;
+
+    for (i = 0; i < 16; i++) {
+        if (block[i * 4 + 3] == 0) {
+            continue;
+        }
+
+        int r = block[i * 4 + 0] - mu[0];
+        int g = block[i * 4 + 1] - mu[1];
+        int b = block[i * 4 + 2] - mu[2];
+
+        cov[0] += r * r;
+        cov[1] += r * g;
+        cov[2] += r * b;
+        cov[3] += g * g;
+        cov[4] += g * b;
+        cov[5] += b * b;
+    }
+
+    // convert covariance matrix to float, find principal axis via power iter
+    for (i = 0; i < 6; i++)
+        covf[i] = static_cast<float>(cov[i]) / 255.0f;
+
+    vfr = (float)(max[0] - min[0]);
+    vfg = (float)(max[1] - min[1]);
+    vfb = (float)(max[2] - min[2]);
+
+    for (iter = 0; iter < nIterPower; iter++) {
+        float r = vfr * covf[0] + vfg * covf[1] + vfb * covf[2];
+        float g = vfr * covf[1] + vfg * covf[3] + vfb * covf[4];
+        float b = vfr * covf[2] + vfg * covf[4] + vfb * covf[5];
+
+        vfr = r;
+        vfg = g;
+        vfb = b;
+    }
+
+    magn = STBD_FABS(vfr);
+    if (STBD_FABS(vfg) > magn)
+        magn = STBD_FABS(vfg);
+    if (STBD_FABS(vfb) > magn)
+        magn = STBD_FABS(vfb);
+
+    if (magn < 4.0f) { // too small, default to luminance
+        v_r = 299;     // JPEG YCbCr luma coefs, scaled by 1000.
+        v_g = 587;
+        v_b = 114;
+    } else {
+        magn = 512.0 / magn;
+        v_r = (int)(vfr * magn);
+        v_g = (int)(vfg * magn);
+        v_b = (int)(vfb * magn);
+    }
+
+    minp = maxp = NULL;
+    mind = 0x7fffffff;
+    maxd = -0x80000000;
+
+    // Pick colors at extreme points
+    for (i = 0; i < 16; i++) {
+        if (block[i * 4 + 3] == 0) {
+            continue;
+        }
+
+        int dot = block[i * 4 + 0] * v_r + block[i * 4 + 1] * v_g + block[i * 4 + 2] * v_b;
+
+        if (dot < mind) {
+            mind = dot;
+            minp = block + i * 4;
+        }
+
+        if (dot > maxd) {
+            maxd = dot;
+            maxp = block + i * 4;
+        }
+    }
+
+    if (!maxp) {
+        // all alpha, no color
+        *pmin16 = 0xffff;
+        *pmax16 = 0;
+    } else {
+        // endpoint colors found
+        *pmax16 = stb__As16Bit(maxp[0], maxp[1], maxp[2]);
+        *pmin16 = stb__As16Bit(minp[0], minp[1], minp[2]);
+
+        if (*pmax16 == *pmin16) {
+            // modify the endpoints to indicate presence of an alpha block
+            if (*pmax16 > 0) {
+                (*pmax16)--;
+            } else {
+                (*pmin16)++;
+            }
+        }
+
+        stb__ReorderColors(pmax16, pmin16);
+    }
+}
+
+static const float stb__midpoints5[32] = {
+    0.015686f, 0.047059f, 0.078431f, 0.111765f, 0.145098f, 0.176471f, 0.207843f, 0.241176f,
+    0.274510f, 0.305882f, 0.337255f, 0.370588f, 0.403922f, 0.435294f, 0.466667f, 0.5f,
+    0.533333f, 0.564706f, 0.596078f, 0.629412f, 0.662745f, 0.694118f, 0.725490f, 0.758824f,
+    0.792157f, 0.823529f, 0.854902f, 0.888235f, 0.921569f, 0.952941f, 0.984314f, 1.0f};
+
+static const float stb__midpoints6[64] = {
+    0.007843f, 0.023529f, 0.039216f, 0.054902f, 0.070588f, 0.086275f, 0.101961f, 0.117647f,
+    0.133333f, 0.149020f, 0.164706f, 0.180392f, 0.196078f, 0.211765f, 0.227451f, 0.245098f,
+    0.262745f, 0.278431f, 0.294118f, 0.309804f, 0.325490f, 0.341176f, 0.356863f, 0.372549f,
+    0.388235f, 0.403922f, 0.419608f, 0.435294f, 0.450980f, 0.466667f, 0.482353f, 0.500000f,
+    0.517647f, 0.533333f, 0.549020f, 0.564706f, 0.580392f, 0.596078f, 0.611765f, 0.627451f,
+    0.643137f, 0.658824f, 0.674510f, 0.690196f, 0.705882f, 0.721569f, 0.737255f, 0.754902f,
+    0.772549f, 0.788235f, 0.803922f, 0.819608f, 0.835294f, 0.850980f, 0.866667f, 0.882353f,
+    0.898039f, 0.913725f, 0.929412f, 0.945098f, 0.960784f, 0.976471f, 0.992157f, 1.0f};
+
+static unsigned short stb__Quantize5(float x) {
+    unsigned short q;
+    x = x < 0 ? 0 : x > 1 ? 1 : x; // saturate
+    q = (unsigned short)(x * 31);
+    q += (x > stb__midpoints5[q]);
+    return q;
+}
+
+static unsigned short stb__Quantize6(float x) {
+    unsigned short q;
+    x = x < 0 ? 0 : x > 1 ? 1 : x; // saturate
+    q = (unsigned short)(x * 63);
+    q += (x > stb__midpoints6[q]);
+    return q;
+}
+
+// The refinement function. (Clever code, part 2)
+// Tries to optimize colors to suit block contents better.
+// (By solving a least squares system via normal equations+Cramer's rule)
+static int stb__RefineBlock(unsigned char* block, unsigned short* pmax16, unsigned short* pmin16,
+                            unsigned int mask) {
+    static const int w1Tab[4] = {3, 0, 2, 1};
+    static const int prods[4] = {0x090000, 0x000900, 0x040102, 0x010402};
+    // ^some magic to save a lot of multiplies in the accumulating loop...
+    // (precomputed products of weights for least squares system, accumulated
+    // inside one 32-bit register)
+
+    float f;
+    unsigned short oldMin, oldMax, min16, max16;
+    int i, akku = 0, xx, xy, yy;
+    int At1_r, At1_g, At1_b;
+    int At2_r, At2_g, At2_b;
+    unsigned int cm = mask;
+
+    oldMin = *pmin16;
+    oldMax = *pmax16;
+
+    if ((mask ^ (mask << 2)) < 4) // all pixels have the same index?
+    {
+        // yes, linear system would be singular; solve using optimal
+        // single-color match on average color
+        int r = 8, g = 8, b = 8;
+        for (i = 0; i < 16; ++i) {
+            r += block[i * 4 + 0];
+            g += block[i * 4 + 1];
+            b += block[i * 4 + 2];
+        }
+
+        r >>= 4;
+        g >>= 4;
+        b >>= 4;
+
+        max16 = static_cast<unsigned short>((stb__OMatch5[r][0] << 11) | (stb__OMatch6[g][0] << 5) |
+                                            stb__OMatch5[b][0]);
+        min16 = static_cast<unsigned short>((stb__OMatch5[r][1] << 11) | (stb__OMatch6[g][1] << 5) |
+                                            stb__OMatch5[b][1]);
+    } else {
+        At1_r = At1_g = At1_b = 0;
+        At2_r = At2_g = At2_b = 0;
+        for (i = 0; i < 16; ++i, cm >>= 2) {
+            int step = cm & 3;
+            int w1 = w1Tab[step];
+            int r = block[i * 4 + 0];
+            int g = block[i * 4 + 1];
+            int b = block[i * 4 + 2];
+
+            akku += prods[step];
+            At1_r += w1 * r;
+            At1_g += w1 * g;
+            At1_b += w1 * b;
+            At2_r += r;
+            At2_g += g;
+            At2_b += b;
+        }
+
+        At2_r = 3 * At2_r - At1_r;
+        At2_g = 3 * At2_g - At1_g;
+        At2_b = 3 * At2_b - At1_b;
+
+        // extract solutions and decide solvability
+        xx = akku >> 16;
+        yy = (akku >> 8) & 0xff;
+        xy = (akku >> 0) & 0xff;
+
+        f = 3.0f / 255.0f / static_cast<float>(xx * yy - xy * xy);
+
+        max16 = static_cast<unsigned short>(
+            stb__Quantize5(static_cast<float>(At1_r * yy - At2_r * xy) * f) << 11);
+        max16 |= static_cast<unsigned short>(
+            stb__Quantize6(static_cast<float>(At1_g * yy - At2_g * xy) * f) << 5);
+        max16 |= static_cast<unsigned short>(
+            stb__Quantize5(static_cast<float>(At1_b * yy - At2_b * xy) * f) << 0);
+
+        min16 = static_cast<unsigned short>(
+            stb__Quantize5(static_cast<float>(At2_r * xx - At1_r * xy) * f) << 11);
+        min16 |= static_cast<unsigned short>(
+            stb__Quantize6(static_cast<float>(At2_g * xx - At1_g * xy) * f) << 5);
+        min16 |= static_cast<unsigned short>(
+            stb__Quantize5(static_cast<float>(At2_b * xx - At1_b * xy) * f) << 0);
+    }
+
+    *pmin16 = min16;
+    *pmax16 = max16;
+    stb__ReorderColors(pmax16, pmin16);
+
+    return oldMin != min16 || oldMax != max16;
+}
+
+// Color block compression
+static void stb__CompressColorBlock(unsigned char* dest, unsigned char* block, int alpha,
+                                    int mode) {
+    unsigned int mask;
+    int i;
+    int refinecount;
+    unsigned short max16, min16;
+    unsigned char color[4 * 4];
+
+    refinecount = (mode & STB_DXT_HIGHQUAL) ? 2 : 1;
+
+    // check if block is constant
+    for (i = 1; i < 16; i++)
+        if (((unsigned int*)block)[i] != ((unsigned int*)block)[0])
+            break;
+
+    if (i == 16 && block[3] == 0 && alpha) { // constant alpha
+        mask = 0xffffffff;
+        max16 = 0;
+        min16 = 0xffff;
+    } else if (i == 16) { // constant color
+        int r = block[0], g = block[1], b = block[2];
+        mask = 0xaaaaaaaa;
+        max16 = static_cast<unsigned short>((stb__OMatch5[r][0] << 11) | (stb__OMatch6[g][0] << 5) |
+                                            stb__OMatch5[b][0]);
+        min16 = static_cast<unsigned short>((stb__OMatch5[r][1] << 11) | (stb__OMatch6[g][1] << 5) |
+                                            stb__OMatch5[b][1]);
+    } else if (alpha) {
+        stb__OptimizeColorsAlphaBlock(block, &max16, &min16);
+        stb__Eval3Colors(color, max16, min16);
+        mask = stb__MatchColorsAlphaBlock(block, color);
+    } else {
+        // first step: PCA+map along principal axis
+        stb__OptimizeColorsBlock(block, &max16, &min16);
+        if (max16 != min16) {
+            stb__Eval4Colors(color, max16, min16);
+            mask = stb__MatchColorsBlock(block, color);
+        } else
+            mask = 0;
+
+        // third step: refine (multiple times if requested)
+        for (i = 0; i < refinecount; i++) {
+            unsigned int lastmask = mask;
+
+            if (stb__RefineBlock(block, &max16, &min16, mask)) {
+                if (max16 != min16) {
+                    stb__Eval4Colors(color, max16, min16);
+                    mask = stb__MatchColorsBlock(block, color);
+                } else {
+                    mask = 0;
+                    break;
+                }
+            }
+
+            if (mask == lastmask)
+                break;
+        }
+    }
+
+    // write the color block
+    if (!alpha)
+        stb__FinalizeColors(&max16, &min16, &mask);
+
+    dest[0] = (unsigned char)(max16);
+    dest[1] = (unsigned char)(max16 >> 8);
+    dest[2] = (unsigned char)(min16);
+    dest[3] = (unsigned char)(min16 >> 8);
+    dest[4] = (unsigned char)(mask);
+    dest[5] = (unsigned char)(mask >> 8);
+    dest[6] = (unsigned char)(mask >> 16);
+    dest[7] = (unsigned char)(mask >> 24);
+}
+
+// Alpha block compression (this is easy for a change)
+static void stb__CompressAlphaBlock(unsigned char* dest, unsigned char* src, int stride) {
+    int i, dist, bias, dist4, dist2, bits, mask;
+
+    // find min/max color
+    int mn, mx;
+    mn = mx = src[0];
+
+    for (i = 1; i < 16; i++) {
+        if (src[i * stride] < mn)
+            mn = src[i * stride];
+        else if (src[i * stride] > mx)
+            mx = src[i * stride];
+    }
+
+    // encode them
+    dest[0] = (unsigned char)mx;
+    dest[1] = (unsigned char)mn;
+    dest += 2;
+
+    // determine bias and emit color indices
+    // given the choice of mx/mn, these indices are optimal:
+    // http://fgiesen.wordpress.com/2009/12/15/dxt5-alpha-block-index-determination/
+    dist = mx - mn;
+    dist4 = dist * 4;
+    dist2 = dist * 2;
+    bias = (dist < 8) ? (dist - 1) : (dist / 2 + 2);
+    bias -= mn * 7;
+    bits = 0, mask = 0;
+
+    for (i = 0; i < 16; i++) {
+        int a = src[i * stride] * 7 + bias;
+        int ind, t;
+
+        // select index. this is a "linear scale" lerp factor between 0 (val=min)
+        // and 7 (val=max).
+        t = (a >= dist4) ? -1 : 0;
+        ind = t & 4;
+        a -= dist4 & t;
+        t = (a >= dist2) ? -1 : 0;
+        ind += t & 2;
+        a -= dist2 & t;
+        ind += (a >= dist);
+
+        // turn linear scale into DXT index (0/1 are extremal pts)
+        ind = -ind & 7;
+        ind ^= (2 > ind);
+
+        // write index
+        mask |= ind << bits;
+        if ((bits += 3) >= 8) {
+            *dest++ = (unsigned char)mask;
+            mask >>= 8;
+            bits -= 8;
+        }
+    }
+}
+
+void stb_compress_bc1_block(unsigned char* dest, const unsigned char* src, int alpha, int mode) {
+    stb__CompressColorBlock(dest, (unsigned char*)src, alpha, mode);
+}
+
+void stb_compress_bc3_block(unsigned char* dest, const unsigned char* src, int mode) {
+    unsigned char data[16][4];
+    int i;
+
+    stb__CompressAlphaBlock(dest, (unsigned char*)src + 3, 4);
+    dest += 8;
+    // make a new copy of the data in which alpha is opaque,
+    // because code uses a fast test for color constancy
+    memcpy(data, src, 4 * 16);
+    for (i = 0; i < 16; ++i)
+        data[i][3] = 255;
+    src = &data[0][0];
+
+    stb__CompressColorBlock(dest, (unsigned char*)src, 0, mode);
+}
@@ -0,0 +1,36 @@
+// SPDX-FileCopyrightText: fabian "ryg" giesen
+// SPDX-License-Identifier: MIT
+
+// stb_dxt.h - v1.12 - DXT1/DXT5 compressor
+
+#ifndef STB_INCLUDE_STB_DXT_H
+#define STB_INCLUDE_STB_DXT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef STB_DXT_STATIC
+#define STBDDEF static
+#else
+#define STBDDEF extern
+#endif
+
+// compression mode (bitflags)
+#define STB_DXT_NORMAL 0
+#define STB_DXT_DITHER 1 // use dithering. was always dubious, now deprecated. does nothing!
+#define STB_DXT_HIGHQUAL                                                                           \
+    2 // high quality mode, does two refinement steps instead of 1. ~30-40% slower.
+
+STBDDEF void stb_compress_bc1_block(unsigned char* dest,
+                                    const unsigned char* src_rgba_four_bytes_per_pixel, int alpha,
+                                    int mode);
+
+STBDDEF void stb_compress_bc3_block(unsigned char* dest, const unsigned char* src, int mode);
+
+#define STB_COMPRESS_DXT_BLOCK
+
+#ifdef __cplusplus
+}
+#endif
+#endif // STB_INCLUDE_STB_DXT_H
@@ -105,7 +105,7 @@ void AudioRenderer::Start(AudioRenderer_Mailbox* mailbox_) {
    }

    mailbox = mailbox_;
-    thread = std::thread(&AudioRenderer::ThreadFunc, this);
+    thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(stop_token); });
    running = true;
 }

@@ -131,7 +131,7 @@ void AudioRenderer::CreateSinkStreams() {
    }
 }

-void AudioRenderer::ThreadFunc() {
+void AudioRenderer::ThreadFunc(std::stop_token stop_token) {
    static constexpr char name[]{"AudioRenderer"};
    MicroProfileOnThreadCreate(name);
    Common::SetCurrentThreadName(name);
@@ -146,7 +146,7 @@ void AudioRenderer::ThreadFunc() {

    constexpr u64 max_process_time{2'304'000ULL};

-    while (true) {
+    while (!stop_token.stop_requested()) {
        auto message{mailbox->ADSPWaitMessage()};
        switch (message) {
        case RenderMessage::AudioRenderer_Shutdown:
@@ -154,6 +154,11 @@ void AudioRenderer::ThreadFunc() {
            return;

        case RenderMessage::AudioRenderer_Render: {
+            if (system.IsShuttingDown()) [[unlikely]] {
+                std::this_thread::sleep_for(std::chrono::milliseconds(5));
+                mailbox->ADSPSendMessage(RenderMessage::AudioRenderer_RenderResponse);
+                continue;
+            }
            std::array<bool, MaxRendererSessions> buffers_reset{};
            std::array<u64, MaxRendererSessions> render_times_taken{};
            const auto start_time{system.CoreTiming().GetClockTicks()};
@@ -189,7 +194,7 @@ void AudioRenderer::ThreadFunc() {
                    max_time = std::min(command_buffer.time_limit, max_time);
                    command_list_processor.SetProcessTimeMax(max_time);

-                    streams[index]->WaitFreeSpace();
+                    streams[index]->WaitFreeSpace(stop_token);

                    // Process the command list
                    {
@@ -177,7 +177,7 @@ private:
    /**
     * Main AudioRenderer thread, responsible for processing the command lists.
     */
-    void ThreadFunc();
+    void ThreadFunc(std::stop_token stop_token);

    /**
     * Creates the streams which will receive the processed samples.
@@ -187,7 +187,7 @@ private:
    /// Core system
    Core::System& system;
    /// Main thread
-    std::thread thread{};
+    std::jthread thread{};
    /// The current state
    std::atomic<bool> running{};
    /// The active mailbox
@@ -27,7 +27,7 @@ bool SystemManager::InitializeUnsafe() {
    if (!active) {
        if (adsp.Start()) {
            active = true;
-            thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(); });
+            thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(stop_token); });
        }
    }

@@ -39,8 +39,7 @@ void SystemManager::Stop() {
        return;
    }
    active = false;
-    update.store(true);
-    update.notify_all();
+    thread.request_stop();
    thread.join();
    adsp.Stop();
 }
@@ -85,12 +84,12 @@ bool SystemManager::Remove(System& system_) {
    return true;
 }

-void SystemManager::ThreadFunc() {
+void SystemManager::ThreadFunc(std::stop_token stop_token) {
    static constexpr char name[]{"AudioRenderSystemManager"};
    MicroProfileOnThreadCreate(name);
    Common::SetCurrentThreadName(name);
    Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
-    while (active) {
+    while (active && !stop_token.stop_requested()) {
        {
            std::scoped_lock l{mutex1};

@@ -66,13 +66,7 @@ private:
    /**
     * Main thread responsible for command generation.
     */
-    void ThreadFunc();
-
-    enum class StreamState {
-        Filling,
-        Steady,
-        Draining,
-    };
+    void ThreadFunc(std::stop_token stop_token);

    /// Core system
    Core::System& core;
@@ -90,8 +84,6 @@ private:
    ADSP::ADSP& adsp;
    /// AudioRenderer mailbox for communication
    ADSP::AudioRenderer_Mailbox* mailbox{};
-    /// Atomic for main thread to wait on
-    std::atomic<bool> update{};
 };

 } // namespace AudioCore::AudioRenderer
@@ -269,10 +269,14 @@ u64 SinkStream::GetExpectedPlayedSampleCount() {
    return std::min<u64>(exp_played_sample_count, max_played_sample_count) + TargetSampleCount * 3;
 }

-void SinkStream::WaitFreeSpace() {
+void SinkStream::WaitFreeSpace(std::stop_token stop_token) {
    std::unique_lock lk{release_mutex};
-    release_cv.wait(
-        lk, [this]() { return queued_buffers < max_queue_size || system.IsShuttingDown(); });
+    release_cv.wait_for(lk, std::chrono::milliseconds(5),
+                        [this]() { return queued_buffers < max_queue_size; });
+    if (queued_buffers > max_queue_size + 3) {
+        Common::CondvarWait(release_cv, lk, stop_token,
+                            [this] { return queued_buffers < max_queue_size; });
+    }
 }

 } // namespace AudioCore::Sink
@@ -13,6 +13,7 @@

 #include "audio_core/common/common.h"
 #include "common/common_types.h"
+#include "common/polyfill_thread.h"
 #include "common/reader_writer_queue.h"
 #include "common/ring_buffer.h"
 #include "common/thread.h"
@@ -210,7 +211,7 @@ public:
    /**
     * Waits for free space in the sample ring buffer
     */
-    void WaitFreeSpace();
+    void WaitFreeSpace(std::stop_token stop_token);

 protected:
    /// Core system
@@ -252,7 +253,7 @@ private:
    /// Set via IAudioDevice service calls
    f32 device_volume{1.0f};
    /// Signalled when ring buffer entries are consumed
-    std::condition_variable release_cv;
+    std::condition_variable_any release_cv;
    std::mutex release_mutex;
 };

@@ -188,3 +188,8 @@ private:

 template <std::size_t Position, std::size_t Bits, typename T>
 using BitFieldBE = BitField<Position, Bits, T, BETag>;
+
+template <std::size_t Position, std::size_t Bits, typename T, typename EndianTag = LETag>
+inline auto format_as(BitField<Position, Bits, T, EndianTag> bitfield) {
+    return bitfield.Value();
+}
@@ -14,6 +14,7 @@
 #ifndef _GNU_SOURCE
 #define _GNU_SOURCE
 #endif
+#include <boost/icl/interval_set.hpp>
 #include <fcntl.h>
 #include <sys/mman.h>
 #include <unistd.h>
@@ -415,6 +416,7 @@ public:
        madvise(virtual_base, virtual_size, MADV_HUGEPAGE);
 #endif

+        placeholders.add({0, virtual_size});
        good = true;
    }

@@ -423,6 +425,10 @@ public:
    }

    void Map(size_t virtual_offset, size_t host_offset, size_t length) {
+        {
+            std::scoped_lock lock{placeholder_mutex};
+            placeholders.subtract({virtual_offset, virtual_offset + length});
+        }

        void* ret = mmap(virtual_base + virtual_offset, length, PROT_READ | PROT_WRITE,
                         MAP_SHARED | MAP_FIXED, fd, host_offset);
@@ -433,6 +439,19 @@ public:
        // The method name is wrong. We're still talking about the virtual range.
        // We don't want to unmap, we want to reserve this memory.

+        {
+            std::scoped_lock lock{placeholder_mutex};
+            auto it = placeholders.find({virtual_offset - 1, virtual_offset + length + 1});
+
+            if (it != placeholders.end()) {
+                size_t prev_upper = virtual_offset + length;
+                virtual_offset = std::min(virtual_offset, it->lower());
+                length = std::max(it->upper(), prev_upper) - virtual_offset;
+            }
+
+            placeholders.add({virtual_offset, virtual_offset + length});
+        }
+
        void* ret = mmap(virtual_base + virtual_offset, length, PROT_NONE,
                         MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
        ASSERT_MSG(ret != MAP_FAILED, "mmap failed: {}", strerror(errno));
@@ -476,6 +495,9 @@ private:
    }

    int fd{-1}; // memfd file descriptor, -1 is the error value of memfd_create
+
+    boost::icl::interval_set<size_t> placeholders; ///< Mapped placeholders
+    std::mutex placeholder_mutex;                  ///< Mutex for placeholders
 };

 #else // ^^^ Linux ^^^ vvv Generic vvv
@@ -61,6 +61,7 @@ void LogSettings() {
    log_setting("Renderer_NvdecEmulation", values.nvdec_emulation.GetValue());
    log_setting("Renderer_AccelerateASTC", values.accelerate_astc.GetValue());
    log_setting("Renderer_AsyncASTC", values.async_astc.GetValue());
+    log_setting("Renderer_AstcRecompression", values.astc_recompression.GetValue());
    log_setting("Renderer_UseVsync", values.vsync_mode.GetValue());
    log_setting("Renderer_UseReactiveFlushing", values.use_reactive_flushing.GetValue());
    log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
@@ -224,6 +225,7 @@ void RestoreGlobalState(bool is_powered_on) {
    values.nvdec_emulation.SetGlobal(true);
    values.accelerate_astc.SetGlobal(true);
    values.async_astc.SetGlobal(true);
+    values.astc_recompression.SetGlobal(true);
    values.use_reactive_flushing.SetGlobal(true);
    values.shader_backend.SetGlobal(true);
    values.use_asynchronous_shaders.SetGlobal(true);
@@ -232,6 +234,7 @@ void RestoreGlobalState(bool is_powered_on) {
    values.bg_red.SetGlobal(true);
    values.bg_green.SetGlobal(true);
    values.bg_blue.SetGlobal(true);
+    values.enable_compute_pipelines.SetGlobal(true);

    // System
    values.language_index.SetGlobal(true);
@@ -90,6 +90,12 @@ enum class AntiAliasing : u32 {
    LastAA = Smaa,
 };

+enum class AstcRecompression : u32 {
+    Uncompressed = 0,
+    Bc1 = 1,
+    Bc3 = 2,
+};
+
 struct ResolutionScalingInfo {
    u32 up_scale{1};
    u32 down_shift{0};
@@ -472,6 +478,10 @@ struct Values {
    SwitchableSetting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
    SwitchableSetting<bool> use_vulkan_driver_pipeline_cache{true,
                                                             "use_vulkan_driver_pipeline_cache"};
+    SwitchableSetting<bool> enable_compute_pipelines{false, "enable_compute_pipelines"};
+    SwitchableSetting<AstcRecompression, true> astc_recompression{
+        AstcRecompression::Uncompressed, AstcRecompression::Uncompressed, AstcRecompression::Bc3,
+        "astc_recompression"};

    SwitchableSetting<u8> bg_red{0, "bg_red"};
    SwitchableSetting<u8> bg_green{0, "bg_green"};
@@ -535,6 +545,8 @@ struct Values {
    Setting<bool> enable_ir_sensor{false, "enable_ir_sensor"};
    Setting<std::string> ir_sensor_device{"auto", "ir_sensor_device"};

+    Setting<bool> random_amiibo_id{false, "random_amiibo_id"};
+
    // Data Storage
    Setting<bool> use_virtual_sd{true, "use_virtual_sd"};
    Setting<bool> gamecard_inserted{false, "gamecard_inserted"};
@@ -106,6 +106,8 @@ add_library(core STATIC
    file_sys/system_archive/time_zone_binary.h
    file_sys/vfs.cpp
    file_sys/vfs.h
+    file_sys/vfs_cached.cpp
+    file_sys/vfs_cached.h
    file_sys/vfs_concat.cpp
    file_sys/vfs_concat.h
    file_sys/vfs_layered.cpp
@@ -117,8 +117,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
            return nullptr;
        }

-        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(std::move(concat),
-                                                                  dir->GetName());
+        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(concat, dir->GetName());
    }

    if (Common::FS::IsDir(path)) {
@@ -21,6 +21,7 @@
 #include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs.h"
+#include "core/file_sys/vfs_cached.h"
 #include "core/file_sys/vfs_layered.h"
 #include "core/file_sys/vfs_vector.h"
 #include "core/hle/service/filesystem/filesystem.h"
@@ -380,11 +381,11 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t

        auto romfs_dir = FindSubdirectoryCaseless(subdir, "romfs");
        if (romfs_dir != nullptr)
-            layers.push_back(std::move(romfs_dir));
+            layers.push_back(std::make_shared<CachedVfsDirectory>(romfs_dir));

        auto ext_dir = FindSubdirectoryCaseless(subdir, "romfs_ext");
        if (ext_dir != nullptr)
-            layers_ext.push_back(std::move(ext_dir));
+            layers_ext.push_back(std::make_shared<CachedVfsDirectory>(ext_dir));
    }

    // When there are no layers to apply, return early as there is no need to rebuild the RomFS
@@ -9,6 +9,7 @@
 #include "core/file_sys/fsmitm_romfsbuild.h"
 #include "core/file_sys/romfs.h"
 #include "core/file_sys/vfs.h"
+#include "core/file_sys/vfs_cached.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/file_sys/vfs_vector.h"
@@ -132,7 +133,7 @@ VirtualDir ExtractRomFS(VirtualFile file, RomFSExtractionType type) {
        out = out->GetSubdirectories().front();
    }

-    return out;
+    return std::make_shared<CachedVfsDirectory>(out);
 }

 VirtualFile CreateRomFS(VirtualDir dir, VirtualDir ext) {
@@ -140,7 +141,8 @@ VirtualFile CreateRomFS(VirtualDir dir, VirtualDir ext) {
        return nullptr;

    RomFSBuildContext ctx{dir, ext};
-    return ConcatenatedVfsFile::MakeConcatenatedFile(0, ctx.Build(), dir->GetName());
+    auto file_map = ctx.Build();
+    return ConcatenatedVfsFile::MakeConcatenatedFile(0, file_map, dir->GetName());
 }

 } // namespace FileSys
@@ -82,9 +82,9 @@ std::string GetFutureSaveDataPath(SaveDataSpaceId space_id, SaveDataType type, u
    // Only detect account/device saves from the future location.
    switch (type) {
    case SaveDataType::SaveData:
-        return fmt::format("{}/account/{}/{:016X}/1", space_id_path, uuid.RawString(), title_id);
+        return fmt::format("{}/account/{}/{:016X}/0", space_id_path, uuid.RawString(), title_id);
    case SaveDataType::DeviceSaveData:
-        return fmt::format("{}/device/{:016X}/1", space_id_path, title_id);
+        return fmt::format("{}/device/{:016X}/0", space_id_path, title_id);
    default:
        return "";
    }
@@ -0,0 +1,63 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/file_sys/vfs_cached.h"
+#include "core/file_sys/vfs_types.h"
+
+namespace FileSys {
+
+CachedVfsDirectory::CachedVfsDirectory(VirtualDir& source_dir)
+    : name(source_dir->GetName()), parent(source_dir->GetParentDirectory()) {
+    for (auto& dir : source_dir->GetSubdirectories()) {
+        dirs.emplace(dir->GetName(), std::make_shared<CachedVfsDirectory>(dir));
+    }
+    for (auto& file : source_dir->GetFiles()) {
+        files.emplace(file->GetName(), file);
+    }
+}
+
+CachedVfsDirectory::~CachedVfsDirectory() = default;
+
+VirtualFile CachedVfsDirectory::GetFile(std::string_view file_name) const {
+    auto it = files.find(file_name);
+    if (it != files.end()) {
+        return it->second;
+    }
+
+    return nullptr;
+}
+
+VirtualDir CachedVfsDirectory::GetSubdirectory(std::string_view dir_name) const {
+    auto it = dirs.find(dir_name);
+    if (it != dirs.end()) {
+        return it->second;
+    }
+
+    return nullptr;
+}
+
+std::vector<VirtualFile> CachedVfsDirectory::GetFiles() const {
+    std::vector<VirtualFile> out;
+    for (auto& [file_name, file] : files) {
+        out.push_back(file);
+    }
+    return out;
+}
+
+std::vector<VirtualDir> CachedVfsDirectory::GetSubdirectories() const {
+    std::vector<VirtualDir> out;
+    for (auto& [dir_name, dir] : dirs) {
+        out.push_back(dir);
+    }
+    return out;
+}
+
+std::string CachedVfsDirectory::GetName() const {
+    return name;
+}
+
+VirtualDir CachedVfsDirectory::GetParentDirectory() const {
+    return parent;
+}
+
+} // namespace FileSys
@@ -0,0 +1,31 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <string_view>
+#include <vector>
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+class CachedVfsDirectory : public ReadOnlyVfsDirectory {
+public:
+    CachedVfsDirectory(VirtualDir& source_directory);
+
+    ~CachedVfsDirectory() override;
+    VirtualFile GetFile(std::string_view file_name) const override;
+    VirtualDir GetSubdirectory(std::string_view dir_name) const override;
+    std::vector<VirtualFile> GetFiles() const override;
+    std::vector<VirtualDir> GetSubdirectories() const override;
+    std::string GetName() const override;
+    VirtualDir GetParentDirectory() const override;
+
+private:
+    std::string name;
+    VirtualDir parent;
+    std::map<std::string, VirtualDir, std::less<>> dirs;
+    std::map<std::string, VirtualFile, std::less<>> files;
+};
+
+} // namespace FileSys
@@ -10,84 +10,105 @@

 namespace FileSys {

-static bool VerifyConcatenationMapContinuity(const std::multimap<u64, VirtualFile>& map) {
-    const auto last_valid = --map.end();
-    for (auto iter = map.begin(); iter != last_valid;) {
-        const auto old = iter++;
-        if (old->first + old->second->GetSize() != iter->first) {
+ConcatenatedVfsFile::ConcatenatedVfsFile(ConcatenationMap&& concatenation_map_, std::string&& name_)
+    : concatenation_map(std::move(concatenation_map_)), name(std::move(name_)) {
+    DEBUG_ASSERT(this->VerifyContinuity());
+}
+
+bool ConcatenatedVfsFile::VerifyContinuity() const {
+    u64 last_offset = 0;
+    for (auto& entry : concatenation_map) {
+        if (entry.offset != last_offset) {
            return false;
        }
+
+        last_offset = entry.offset + entry.file->GetSize();
    }

-    return map.begin()->first == 0;
-}
-
-ConcatenatedVfsFile::ConcatenatedVfsFile(std::vector<VirtualFile> files_, std::string name_)
-    : name(std::move(name_)) {
-    std::size_t next_offset = 0;
-    for (const auto& file : files_) {
-        files.emplace(next_offset, file);
-        next_offset += file->GetSize();
-    }
-}
-
-ConcatenatedVfsFile::ConcatenatedVfsFile(std::multimap<u64, VirtualFile> files_, std::string name_)
-    : files(std::move(files_)), name(std::move(name_)) {
-    ASSERT(VerifyConcatenationMapContinuity(files));
+    return true;
 }

 ConcatenatedVfsFile::~ConcatenatedVfsFile() = default;

-VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(std::vector<VirtualFile> files,
-                                                      std::string name) {
-    if (files.empty())
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(const std::vector<VirtualFile>& files,
+                                                      std::string&& name) {
+    // Fold trivial cases.
+    if (files.empty()) {
        return nullptr;
-    if (files.size() == 1)
-        return files[0];
+    }
+    if (files.size() == 1) {
+        return files.front();
+    }

-    return VirtualFile(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+    // Make the concatenation map from the input.
+    std::vector<ConcatenationEntry> concatenation_map;
+    concatenation_map.reserve(files.size());
+    u64 last_offset = 0;
+
+    for (auto& file : files) {
+        concatenation_map.emplace_back(ConcatenationEntry{
+            .offset = last_offset,
+            .file = file,
+        });
+
+        last_offset += file->GetSize();
+    }
+
+    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
 }

 VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
-                                                      std::multimap<u64, VirtualFile> files,
-                                                      std::string name) {
-    if (files.empty())
+                                                      const std::multimap<u64, VirtualFile>& files,
+                                                      std::string&& name) {
+    // Fold trivial cases.
+    if (files.empty()) {
        return nullptr;
-    if (files.size() == 1)
+    }
+    if (files.size() == 1) {
        return files.begin()->second;
-
-    const auto last_valid = --files.end();
-    for (auto iter = files.begin(); iter != last_valid;) {
-        const auto old = iter++;
-        if (old->first + old->second->GetSize() != iter->first) {
-            files.emplace(old->first + old->second->GetSize(),
-                          std::make_shared<StaticVfsFile>(filler_byte, iter->first - old->first -
-                                                                           old->second->GetSize()));
-        }
    }

-    // Ensure the map starts at offset 0 (start of file), otherwise pad to fill.
-    if (files.begin()->first != 0)
-        files.emplace(0, std::make_shared<StaticVfsFile>(filler_byte, files.begin()->first));
+    // Make the concatenation map from the input.
+    std::vector<ConcatenationEntry> concatenation_map;

-    return VirtualFile(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+    concatenation_map.reserve(files.size());
+    u64 last_offset = 0;
+
+    // Iteration of a multimap is ordered, so offset will be strictly non-decreasing.
+    for (auto& [offset, file] : files) {
+        if (offset > last_offset) {
+            concatenation_map.emplace_back(ConcatenationEntry{
+                .offset = last_offset,
+                .file = std::make_shared<StaticVfsFile>(filler_byte, offset - last_offset),
+            });
+        }
+
+        concatenation_map.emplace_back(ConcatenationEntry{
+            .offset = offset,
+            .file = file,
+        });
+
+        last_offset = offset + file->GetSize();
+    }
+
+    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
 }

 std::string ConcatenatedVfsFile::GetName() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
        return "";
    }
    if (!name.empty()) {
        return name;
    }
-    return files.begin()->second->GetName();
+    return concatenation_map.front().file->GetName();
 }

 std::size_t ConcatenatedVfsFile::GetSize() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
        return 0;
    }
-    return files.rbegin()->first + files.rbegin()->second->GetSize();
+    return concatenation_map.back().offset + concatenation_map.back().file->GetSize();
 }

 bool ConcatenatedVfsFile::Resize(std::size_t new_size) {
@@ -95,10 +116,10 @@ bool ConcatenatedVfsFile::Resize(std::size_t new_size) {
 }

 VirtualDir ConcatenatedVfsFile::GetContainingDirectory() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
        return nullptr;
    }
-    return files.begin()->second->GetContainingDirectory();
+    return concatenation_map.front().file->GetContainingDirectory();
 }

 bool ConcatenatedVfsFile::IsWritable() const {
@@ -110,25 +131,45 @@ bool ConcatenatedVfsFile::IsReadable() const {
 }

 std::size_t ConcatenatedVfsFile::Read(u8* data, std::size_t length, std::size_t offset) const {
-    auto entry = --files.end();
-    for (auto iter = files.begin(); iter != files.end(); ++iter) {
-        if (iter->first > offset) {
-            entry = --iter;
+    const ConcatenationEntry key{
+        .offset = offset,
+        .file = nullptr,
+    };
+
+    // Read nothing if the map is empty.
+    if (concatenation_map.empty()) {
+        return 0;
+    }
+
+    // Binary search to find the iterator to the first position we can check.
+    // It must exist, since we are not empty and are comparing unsigned integers.
+    auto it = std::prev(std::upper_bound(concatenation_map.begin(), concatenation_map.end(), key));
+    u64 cur_length = length;
+    u64 cur_offset = offset;
+
+    while (cur_length > 0 && it != concatenation_map.end()) {
+        // Check if we can read the file at this position.
+        const auto& file = it->file;
+        const u64 file_offset = it->offset;
+        const u64 file_size = file->GetSize();
+
+        if (cur_offset >= file_offset + file_size) {
+            // Entirely out of bounds read.
            break;
        }
+
+        // Read the file at this position.
+        const u64 intended_read_size = std::min<u64>(cur_length, file_size);
+        const u64 actual_read_size =
+            file->Read(data + (cur_offset - offset), intended_read_size, cur_offset - file_offset);
+
+        // Update tracking.
+        cur_offset += actual_read_size;
+        cur_length -= actual_read_size;
+        it++;
    }

-    if (entry->first + entry->second->GetSize() <= offset)
-        return 0;
-
-    const auto read_in =
-        std::min<u64>(entry->first + entry->second->GetSize() - offset, entry->second->GetSize());
-    if (length > read_in) {
-        return entry->second->Read(data, read_in, offset - entry->first) +
-               Read(data + read_in, length - read_in, offset + read_in);
-    }
-
-    return entry->second->Read(data, std::min<u64>(read_in, length), offset - entry->first);
+    return cur_offset - offset;
 }

 std::size_t ConcatenatedVfsFile::Write(const u8* data, std::size_t length, std::size_t offset) {
@@ -3,6 +3,7 @@

 #pragma once

+#include <compare>
 #include <map>
 #include <memory>
 #include "core/file_sys/vfs.h"
@@ -12,19 +13,33 @@ namespace FileSys {
 // Class that wraps multiple vfs files and concatenates them, making reads seamless. Currently
 // read-only.
 class ConcatenatedVfsFile : public VfsFile {
-    explicit ConcatenatedVfsFile(std::vector<VirtualFile> files, std::string name_);
-    explicit ConcatenatedVfsFile(std::multimap<u64, VirtualFile> files, std::string name_);
+private:
+    struct ConcatenationEntry {
+        u64 offset;
+        VirtualFile file;
+
+        auto operator<=>(const ConcatenationEntry& other) const {
+            return this->offset <=> other.offset;
+        }
+    };
+    using ConcatenationMap = std::vector<ConcatenationEntry>;
+
+    explicit ConcatenatedVfsFile(std::vector<ConcatenationEntry>&& concatenation_map,
+                                 std::string&& name);
+    bool VerifyContinuity() const;

 public:
    ~ConcatenatedVfsFile() override;

    /// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
-    static VirtualFile MakeConcatenatedFile(std::vector<VirtualFile> files, std::string name);
+    static VirtualFile MakeConcatenatedFile(const std::vector<VirtualFile>& files,
+                                            std::string&& name);

    /// Convenience function that turns a map of offsets to files into a concatenated file, filling
    /// gaps with a given filler byte.
-    static VirtualFile MakeConcatenatedFile(u8 filler_byte, std::multimap<u64, VirtualFile> files,
-                                            std::string name);
+    static VirtualFile MakeConcatenatedFile(u8 filler_byte,
+                                            const std::multimap<u64, VirtualFile>& files,
+                                            std::string&& name);

    std::string GetName() const override;
    std::size_t GetSize() const override;
@@ -37,8 +52,7 @@ public:
    bool Rename(std::string_view new_name) override;

 private:
-    // Maps starting offset to file -- more efficient.
-    std::multimap<u64, VirtualFile> files;
+    ConcatenationMap concatenation_map;
    std::string name;
 };

@@ -67,23 +67,6 @@ VectorVfsDirectory::VectorVfsDirectory(std::vector<VirtualFile> files_,

 VectorVfsDirectory::~VectorVfsDirectory() = default;

-VirtualFile VectorVfsDirectory::GetFile(std::string_view file_name) const {
-    if (!optimized_file_index_built) {
-        optimized_file_index.clear();
-        for (size_t i = 0; i < files.size(); i++) {
-            optimized_file_index.emplace(files[i]->GetName(), i);
-        }
-        optimized_file_index_built = true;
-    }
-
-    const auto it = optimized_file_index.find(file_name);
-    if (it != optimized_file_index.end()) {
-        return files[it->second];
-    }
-
-    return nullptr;
-}
-
 std::vector<VirtualFile> VectorVfsDirectory::GetFiles() const {
    return files;
 }
@@ -124,7 +107,6 @@ bool VectorVfsDirectory::DeleteSubdirectory(std::string_view subdir_name) {
 }

 bool VectorVfsDirectory::DeleteFile(std::string_view file_name) {
-    optimized_file_index_built = false;
    return FindAndRemoveVectorElement(files, file_name);
 }

@@ -142,7 +124,6 @@ VirtualFile VectorVfsDirectory::CreateFile(std::string_view file_name) {
 }

 void VectorVfsDirectory::AddFile(VirtualFile file) {
-    optimized_file_index_built = false;
    files.push_back(std::move(file));
 }

@@ -105,7 +105,6 @@ public:
                                VirtualDir parent = nullptr);
    ~VectorVfsDirectory() override;

-    VirtualFile GetFile(std::string_view file_name) const override;
    std::vector<VirtualFile> GetFiles() const override;
    std::vector<VirtualDir> GetSubdirectories() const override;
    bool IsWritable() const override;
@@ -127,9 +126,6 @@ private:

    VirtualDir parent;
    std::string name;
-
-    mutable std::map<std::string, size_t, std::less<>> optimized_file_index;
-    mutable bool optimized_file_index_built{};
 };

 } // namespace FileSys
@@ -1283,9 +1283,14 @@ bool EmulatedController::HasNfc() const {
 }

 bool EmulatedController::WriteNfc(const std::vector<u8>& data) {
-    auto& nfc_output_device = output_devices[3];
+    auto& nfc_output_device = output_devices[static_cast<std::size_t>(DeviceIndex::Right)];
+    auto& nfc_virtual_output_device = output_devices[3];

-    return nfc_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
+    if (nfc_output_device->SupportsNfc() != Common::Input::NfcState::NotSupported) {
+        return nfc_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
+    }
+
+    return nfc_virtual_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
 }

 void EmulatedController::SetLedPattern() {
@@ -144,14 +144,10 @@ private:

 class KScopedMemoryBlockManagerAuditor {
 public:
-    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager* m) : m_manager(m) {
-        ASSERT(m_manager->CheckState());
-    }
+    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager* m) : m_manager(m) {}
    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager& m)
        : KScopedMemoryBlockManagerAuditor(std::addressof(m)) {}
-    ~KScopedMemoryBlockManagerAuditor() {
-        ASSERT(m_manager->CheckState());
-    }
+    ~KScopedMemoryBlockManagerAuditor() = default;

 private:
    KMemoryBlockManager* m_manager;
@@ -13,6 +13,7 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_transfer_memory.h"
+#include "core/hle/result.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applet_ae.h"
@@ -1335,7 +1336,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
        {24, nullptr, "GetLaunchStorageInfoForDebug"},
        {25, &IApplicationFunctions::ExtendSaveData, "ExtendSaveData"},
        {26, &IApplicationFunctions::GetSaveDataSize, "GetSaveDataSize"},
-        {27, nullptr, "CreateCacheStorage"},
+        {27, &IApplicationFunctions::CreateCacheStorage, "CreateCacheStorage"},
        {28, nullptr, "GetSaveDataSizeMax"},
        {29, nullptr, "GetCacheStorageMax"},
        {30, &IApplicationFunctions::BeginBlockingHomeButtonShortAndLongPressed, "BeginBlockingHomeButtonShortAndLongPressed"},
@@ -1738,6 +1739,36 @@ void IApplicationFunctions::GetSaveDataSize(HLERequestContext& ctx) {
    rb.Push(size.journal);
 }

+void IApplicationFunctions::CreateCacheStorage(HLERequestContext& ctx) {
+    struct InputParameters {
+        u16 index;
+        s64 size;
+        s64 journal_size;
+    };
+    static_assert(sizeof(InputParameters) == 24);
+
+    struct OutputParameters {
+        u32 storage_target;
+        u64 required_size;
+    };
+    static_assert(sizeof(OutputParameters) == 16);
+
+    IPC::RequestParser rp{ctx};
+    const auto params = rp.PopRaw<InputParameters>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called with index={}, size={:#x}, journal_size={:#x}",
+                params.index, params.size, params.journal_size);
+
+    const OutputParameters resp{
+        .storage_target = 1,
+        .required_size = 0,
+    };
+
+    IPC::ResponseBuilder rb{ctx, 6};
+    rb.Push(ResultSuccess);
+    rb.PushRaw(resp);
+}
+
 void IApplicationFunctions::QueryApplicationPlayStatistics(HLERequestContext& ctx) {
    LOG_WARNING(Service_AM, "(STUBBED) called");

@@ -333,6 +333,7 @@ private:
    void GetPseudoDeviceId(HLERequestContext& ctx);
    void ExtendSaveData(HLERequestContext& ctx);
    void GetSaveDataSize(HLERequestContext& ctx);
+    void CreateCacheStorage(HLERequestContext& ctx);
    void BeginBlockingHomeButtonShortAndLongPressed(HLERequestContext& ctx);
    void EndBlockingHomeButtonShortAndLongPressed(HLERequestContext& ctx);
    void BeginBlockingHomeButton(HLERequestContext& ctx);
@@ -24,8 +24,10 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/system_archive/system_archive.h"
 #include "core/file_sys/vfs.h"
+#include "core/hle/result.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/filesystem/fsp_srv.h"
+#include "core/hle/service/hle_ipc.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/reporter.h"

@@ -552,9 +554,9 @@ public:
        // Write the data to memory
        ctx.WriteBuffer(begin, range_size);

-        IPC::ResponseBuilder rb{ctx, 3};
+        IPC::ResponseBuilder rb{ctx, 4};
        rb.Push(ResultSuccess);
-        rb.Push<u32>(static_cast<u32>(actual_entries));
+        rb.Push<u64>(actual_entries);
    }

 private:
@@ -712,7 +714,7 @@ FSP_SRV::FSP_SRV(Core::System& system_)
        {59, nullptr, "WriteSaveDataFileSystemExtraData"},
        {60, nullptr, "OpenSaveDataInfoReader"},
        {61, &FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId, "OpenSaveDataInfoReaderBySaveDataSpaceId"},
-        {62, nullptr, "OpenCacheStorageList"},
+        {62, &FSP_SRV::OpenSaveDataInfoReaderOnlyCacheStorage, "OpenSaveDataInfoReaderOnlyCacheStorage"},
        {64, nullptr, "OpenSaveDataInternalStorageFileSystem"},
        {65, nullptr, "UpdateSaveDataMacForDebug"},
        {66, nullptr, "WriteSaveDataFileSystemExtraData2"},
@@ -921,6 +923,15 @@ void FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId(HLERequestContext& ctx) {
        std::make_shared<ISaveDataInfoReader>(system, space, fsc));
 }

+void FSP_SRV::OpenSaveDataInfoReaderOnlyCacheStorage(HLERequestContext& ctx) {
+    LOG_WARNING(Service_FS, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(ResultSuccess);
+    rb.PushIpcInterface<ISaveDataInfoReader>(system, FileSys::SaveDataSpaceId::TemporaryStorage,
+                                             fsc);
+}
+
 void FSP_SRV::WriteSaveDataFileSystemExtraDataBySaveDataAttribute(HLERequestContext& ctx) {
    LOG_WARNING(Service_FS, "(STUBBED) called.");

@@ -957,16 +968,20 @@ void FSP_SRV::ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute(HLERequ
 void FSP_SRV::OpenDataStorageByCurrentProcess(HLERequestContext& ctx) {
    LOG_DEBUG(Service_FS, "called");

-    auto current_romfs = fsc.OpenRomFSCurrentProcess();
-    if (current_romfs.Failed()) {
-        // TODO (bunnei): Find the right error code to use here
-        LOG_CRITICAL(Service_FS, "no file system interface available!");
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultUnknown);
-        return;
+    if (!romfs) {
+        auto current_romfs = fsc.OpenRomFSCurrentProcess();
+        if (current_romfs.Failed()) {
+            // TODO (bunnei): Find the right error code to use here
+            LOG_CRITICAL(Service_FS, "no file system interface available!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        romfs = current_romfs.Unwrap();
    }

-    auto storage = std::make_shared<IStorage>(system, std::move(current_romfs.Unwrap()));
+    auto storage = std::make_shared<IStorage>(system, romfs);

    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
    rb.Push(ResultSuccess);
@@ -42,6 +42,7 @@ private:
    void OpenSaveDataFileSystem(HLERequestContext& ctx);
    void OpenReadOnlySaveDataFileSystem(HLERequestContext& ctx);
    void OpenSaveDataInfoReaderBySaveDataSpaceId(HLERequestContext& ctx);
+    void OpenSaveDataInfoReaderOnlyCacheStorage(HLERequestContext& ctx);
    void WriteSaveDataFileSystemExtraDataBySaveDataAttribute(HLERequestContext& ctx);
    void ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute(HLERequestContext& ctx);
    void OpenDataStorageByCurrentProcess(HLERequestContext& ctx);
@@ -979,8 +979,8 @@ void Controller_NPad::VibrateController(
 }

 void Controller_NPad::VibrateControllers(
-    const std::vector<Core::HID::VibrationDeviceHandle>& vibration_device_handles,
-    const std::vector<Core::HID::VibrationValue>& vibration_values) {
+    std::span<const Core::HID::VibrationDeviceHandle> vibration_device_handles,
+    std::span<const Core::HID::VibrationValue> vibration_values) {
    if (!Settings::values.vibration_enabled.GetValue() && !permit_vibration_session_enabled) {
        return;
    }
@@ -112,8 +112,8 @@ public:
                           const Core::HID::VibrationValue& vibration_value);

    void VibrateControllers(
-        const std::vector<Core::HID::VibrationDeviceHandle>& vibration_device_handles,
-        const std::vector<Core::HID::VibrationValue>& vibration_values);
+        std::span<const Core::HID::VibrationDeviceHandle> vibration_device_handles,
+        std::span<const Core::HID::VibrationValue> vibration_values);

    Core::HID::VibrationValue GetLastVibration(
        const Core::HID::VibrationDeviceHandle& vibration_device_handle) const;
@@ -1601,16 +1601,16 @@ void Hid::SendVibrationValues(HLERequestContext& ctx) {
    IPC::RequestParser rp{ctx};
    const auto applet_resource_user_id{rp.Pop<u64>()};

-    const auto handles = ctx.ReadBuffer(0);
-    const auto vibrations = ctx.ReadBuffer(1);
+    const auto handle_data = ctx.ReadBuffer(0);
+    const auto handle_count = ctx.GetReadBufferNumElements<Core::HID::VibrationDeviceHandle>(0);
+    const auto vibration_data = ctx.ReadBuffer(1);
+    const auto vibration_count = ctx.GetReadBufferNumElements<Core::HID::VibrationValue>(1);

-    std::vector<Core::HID::VibrationDeviceHandle> vibration_device_handles(
-        handles.size() / sizeof(Core::HID::VibrationDeviceHandle));
-    std::vector<Core::HID::VibrationValue> vibration_values(vibrations.size() /
-                                                            sizeof(Core::HID::VibrationValue));
-
-    std::memcpy(vibration_device_handles.data(), handles.data(), handles.size());
-    std::memcpy(vibration_values.data(), vibrations.data(), vibrations.size());
+    auto vibration_device_handles =
+        std::span(reinterpret_cast<const Core::HID::VibrationDeviceHandle*>(handle_data.data()),
+                  handle_count);
+    auto vibration_values = std::span(
+        reinterpret_cast<const Core::HID::VibrationValue*>(vibration_data.data()), vibration_count);

    applet_resource->GetController<Controller_NPad>(HidController::NPad)
        .VibrateControllers(vibration_device_handles, vibration_values);
@@ -52,9 +52,6 @@ bool IsAmiiboValid(const EncryptedNTAG215File& ntag_file) {
    if (ntag_file.compability_container != 0xEEFF10F1U) {
        return false;
    }
-    if (amiibo_data.constant_value != 0xA5) {
-        return false;
-    }
    if (amiibo_data.model_info.tag_type != NFC::PackedTagType::Type2) {
        return false;
    }
@@ -48,9 +48,6 @@ NfcDevice::NfcDevice(Core::HID::NpadIdType npad_id_, Core::System& system_,
    };
    is_controller_set = true;
    callback_key = npad_device->SetCallback(engine_callback);
-
-    auto& standard_steady_clock{system.GetTimeManager().GetStandardSteadyClockCore()};
-    current_posix_time = standard_steady_clock.GetCurrentTimePoint(system).time_point;
 }

 NfcDevice::~NfcDevice() {
@@ -122,18 +119,31 @@ bool NfcDevice::LoadNfcTag(std::span<const u8> data) {

    memcpy(&tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
    is_plain_amiibo = NFP::AmiiboCrypto::IsAmiiboValid(tag_data);
-
-    if (is_plain_amiibo) {
-        encrypted_tag_data = NFP::AmiiboCrypto::EncodedDataToNfcData(tag_data);
-        LOG_INFO(Service_NFP, "Using plain amiibo");
-    } else {
-        tag_data = {};
-        memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
-    }
+    is_write_protected = false;

    device_state = DeviceState::TagFound;
    deactivate_event->GetReadableEvent().Clear();
    activate_event->Signal();
+
+    // Fallback for plain amiibos
+    if (is_plain_amiibo) {
+        LOG_INFO(Service_NFP, "Using plain amiibo");
+        encrypted_tag_data = NFP::AmiiboCrypto::EncodedDataToNfcData(tag_data);
+        return true;
+    }
+
+    // Fallback for encrypted amiibos without keys
+    if (!NFP::AmiiboCrypto::IsKeyAvailable()) {
+        LOG_INFO(Service_NFC, "Loading amiibo without keys");
+        memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
+        BuildAmiiboWithoutKeys();
+        is_plain_amiibo = true;
+        is_write_protected = true;
+        return true;
+    }
+
+    tag_data = {};
+    memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
    return true;
 }

@@ -227,11 +237,21 @@ Result NfcDevice::GetTagInfo(NFP::TagInfo& tag_info, bool is_mifare) const {
        return ResultWrongDeviceState;
    }

+    UniqueSerialNumber uuid = encrypted_tag_data.uuid.uid;
+
+    // Generate random UUID to bypass amiibo load limits
+    if (Settings::values.random_amiibo_id) {
+        Common::TinyMT rng{};
+        rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
+        rng.GenerateRandomBytes(uuid.data(), sizeof(UniqueSerialNumber));
+        uuid[3] = 0x88 ^ uuid[0] ^ uuid[1] ^ uuid[2];
+    }
+
    if (is_mifare) {
        tag_info = {
-            .uuid = encrypted_tag_data.uuid.uid,
+            .uuid = uuid,
            .uuid_extension = {},
-            .uuid_length = static_cast<u8>(encrypted_tag_data.uuid.uid.size()),
+            .uuid_length = static_cast<u8>(uuid.size()),
            .protocol = NfcProtocol::TypeA,
            .tag_type = TagType::Type4,
        };
@@ -240,9 +260,9 @@ Result NfcDevice::GetTagInfo(NFP::TagInfo& tag_info, bool is_mifare) const {

    // Protocol and tag type may change here
    tag_info = {
-        .uuid = encrypted_tag_data.uuid.uid,
+        .uuid = uuid,
        .uuid_extension = {},
-        .uuid_length = static_cast<u8>(encrypted_tag_data.uuid.uid.size()),
+        .uuid_length = static_cast<u8>(uuid.size()),
        .protocol = NfcProtocol::TypeA,
        .tag_type = TagType::Type2,
    };
@@ -339,23 +359,15 @@ Result NfcDevice::Mount(NFP::ModelType model_type, NFP::MountTarget mount_target
        return ResultWrongDeviceState;
    }

-    // The loaded amiibo is not encrypted
-    if (is_plain_amiibo) {
-        device_state = DeviceState::TagMounted;
-        mount_target = mount_target_;
-        return ResultSuccess;
-    }
-
    if (!NFP::AmiiboCrypto::IsAmiiboValid(encrypted_tag_data)) {
        LOG_ERROR(Service_NFP, "Not an amiibo");
        return ResultNotAnAmiibo;
    }

-    // Mark amiibos as read only when keys are missing
-    if (!NFP::AmiiboCrypto::IsKeyAvailable()) {
-        LOG_ERROR(Service_NFP, "No keys detected");
+    // The loaded amiibo is not encrypted
+    if (is_plain_amiibo) {
        device_state = DeviceState::TagMounted;
-        mount_target = NFP::MountTarget::Rom;
+        mount_target = mount_target_;
        return ResultSuccess;
    }

@@ -406,7 +418,7 @@ Result NfcDevice::Flush() {

    auto& settings = tag_data.settings;

-    const auto& current_date = GetAmiiboDate(current_posix_time);
+    const auto& current_date = GetAmiiboDate(GetCurrentPosixTime());
    if (settings.write_date.raw_date != current_date.raw_date) {
        settings.write_date = current_date;
        UpdateSettingsCrc();
@@ -414,11 +426,11 @@ Result NfcDevice::Flush() {

    tag_data.write_counter++;

-    FlushWithBreak(NFP::BreakType::Normal);
+    const auto result = FlushWithBreak(NFP::BreakType::Normal);

    is_data_moddified = false;

-    return ResultSuccess;
+    return result;
 }

 Result NfcDevice::FlushDebug() {
@@ -437,11 +449,11 @@ Result NfcDevice::FlushDebug() {

    tag_data.write_counter++;

-    FlushWithBreak(NFP::BreakType::Normal);
+    const auto result = FlushWithBreak(NFP::BreakType::Normal);

    is_data_moddified = false;

-    return ResultSuccess;
+    return result;
 }

 Result NfcDevice::FlushWithBreak(NFP::BreakType break_type) {
@@ -450,6 +462,11 @@ Result NfcDevice::FlushWithBreak(NFP::BreakType break_type) {
        return ResultWrongDeviceState;
    }

+    if (is_write_protected) {
+        LOG_ERROR(Service_NFP, "No keys available skipping write request");
+        return ResultSuccess;
+    }
+
    std::vector<u8> data(sizeof(NFP::EncryptedNTAG215File));
    if (is_plain_amiibo) {
        memcpy(data.data(), &tag_data, sizeof(tag_data));
@@ -525,6 +542,7 @@ Result NfcDevice::GetModelInfo(NFP::ModelInfo& model_info) const {
    }

    const auto& model_info_data = encrypted_tag_data.user_memory.model_info;
+
    model_info = {
        .character_id = model_info_data.character_id,
        .character_variant = model_info_data.character_variant,
@@ -669,6 +687,7 @@ Result NfcDevice::DeleteRegisterInfo() {
    }

    Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
    rng.GenerateRandomBytes(&tag_data.owner_mii, sizeof(tag_data.owner_mii));
    rng.GenerateRandomBytes(&tag_data.settings.amiibo_name, sizeof(tag_data.settings.amiibo_name));
    rng.GenerateRandomBytes(&tag_data.unknown, sizeof(u8));
@@ -701,7 +720,7 @@ Result NfcDevice::SetRegisterInfoPrivate(const NFP::RegisterInfoPrivate& registe
    auto& settings = tag_data.settings;

    if (tag_data.settings.settings.amiibo_initialized == 0) {
-        settings.init_date = GetAmiiboDate(current_posix_time);
+        settings.init_date = GetAmiiboDate(GetCurrentPosixTime());
        settings.write_date.raw_date = 0;
    }

@@ -868,6 +887,7 @@ Result NfcDevice::SetApplicationArea(std::span<const u8> data) {
    }

    Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
    std::memcpy(tag_data.application_area.data(), data.data(), data.size());
    // Fill remaining data with random numbers
    rng.GenerateRandomBytes(tag_data.application_area.data() + data.size(),
@@ -924,6 +944,7 @@ Result NfcDevice::RecreateApplicationArea(u32 access_id, std::span<const u8> dat
    }

    Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
    std::memcpy(tag_data.application_area.data(), data.data(), data.size());
    // Fill remaining data with random numbers
    rng.GenerateRandomBytes(tag_data.application_area.data() + data.size(),
@@ -973,6 +994,7 @@ Result NfcDevice::DeleteApplicationArea() {
    }

    Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
    rng.GenerateRandomBytes(tag_data.application_area.data(), sizeof(NFP::ApplicationArea));
    rng.GenerateRandomBytes(&tag_data.application_id, sizeof(u64));
    rng.GenerateRandomBytes(&tag_data.application_area_id, sizeof(u32));
@@ -1021,7 +1043,6 @@ Result NfcDevice::GetAll(NFP::NfpData& data) const {
    }

    NFP::CommonInfo common_info{};
-    Service::Mii::MiiManager manager;
    const u64 application_id = tag_data.application_id;

    GetCommonInfo(common_info);
@@ -1189,6 +1210,11 @@ NFP::AmiiboDate NfcDevice::GetAmiiboDate(s64 posix_time) const {
    return amiibo_date;
 }

+u64 NfcDevice::GetCurrentPosixTime() const {
+    auto& standard_steady_clock{system.GetTimeManager().GetStandardSteadyClockCore()};
+    return standard_steady_clock.GetCurrentTimePoint(system).time_point;
+}
+
 u64 NfcDevice::RemoveVersionByte(u64 application_id) const {
    return application_id & ~(0xfULL << NFP::application_id_version_offset);
 }
@@ -1232,6 +1258,28 @@ void NfcDevice::UpdateRegisterInfoCrc() {
    tag_data.register_info_crc = crc.checksum();
 }

+void NfcDevice::BuildAmiiboWithoutKeys() {
+    Service::Mii::MiiManager manager;
+    auto& settings = tag_data.settings;
+
+    tag_data = NFP::AmiiboCrypto::NfcDataToEncodedData(encrypted_tag_data);
+
+    // Common info
+    tag_data.write_counter = 0;
+    tag_data.amiibo_version = 0;
+    settings.write_date = GetAmiiboDate(GetCurrentPosixTime());
+
+    // Register info
+    SetAmiiboName(settings, {'y', 'u', 'z', 'u', 'A', 'm', 'i', 'i', 'b', 'o'});
+    settings.settings.font_region.Assign(0);
+    settings.init_date = GetAmiiboDate(GetCurrentPosixTime());
+    tag_data.owner_mii = manager.BuildFromStoreData(manager.BuildDefault(0));
+
+    // Admin info
+    settings.settings.amiibo_initialized.Assign(1);
+    settings.settings.appdata_initialized.Assign(0);
+}
+
 u64 NfcDevice::GetHandle() const {
    // Generate a handle based of the npad id
    return static_cast<u64>(npad_id);
@@ -105,10 +105,13 @@ private:
    NFP::AmiiboName GetAmiiboName(const NFP::AmiiboSettings& settings) const;
    void SetAmiiboName(NFP::AmiiboSettings& settings, const NFP::AmiiboName& amiibo_name);
    NFP::AmiiboDate GetAmiiboDate(s64 posix_time) const;
+    u64 GetCurrentPosixTime() const;
    u64 RemoveVersionByte(u64 application_id) const;
    void UpdateSettingsCrc();
    void UpdateRegisterInfoCrc();

+    void BuildAmiiboWithoutKeys();
+
    bool is_controller_set{};
    int callback_key;
    const Core::HID::NpadIdType npad_id;
@@ -127,7 +130,7 @@ private:
    bool is_data_moddified{};
    bool is_app_area_open{};
    bool is_plain_amiibo{};
-    s64 current_posix_time{};
+    bool is_write_protected{};
    NFP::MountTarget mount_target{NFP::MountTarget::None};

    NFP::NTAG215File tag_data{};
@@ -64,7 +64,7 @@ void SyncpointManager::FreeSyncpoint(u32 id) {
 }

 bool SyncpointManager::IsSyncpointAllocated(u32 id) const {
-    return (id <= SyncpointCount) && syncpoints[id].reserved;
+    return (id < SyncpointCount) && syncpoints[id].reserved;
 }

 bool SyncpointManager::HasSyncpointExpired(u32 id, u32 threshold) const {
@@ -793,6 +793,7 @@ Status BufferQueueProducer::SetPreallocatedBuffer(s32 slot,
    std::scoped_lock lock{core->mutex};

    slots[slot] = {};
+    slots[slot].fence = Fence::NoFence();
    slots[slot].graphic_buffer = buffer;
    slots[slot].frame_number = 0;

@@ -854,7 +855,7 @@ void BufferQueueProducer::Transact(HLERequestContext& ctx, TransactionId code, u
        status = DequeueBuffer(&slot, &fence, is_async, width, height, pixel_format, usage);

        parcel_out.Write(slot);
-        parcel_out.WriteObject(&fence);
+        parcel_out.WriteFlattenedObject(&fence);
        break;
    }
    case TransactionId::RequestBuffer: {
@@ -864,7 +865,7 @@ void BufferQueueProducer::Transact(HLERequestContext& ctx, TransactionId code, u

        status = RequestBuffer(slot, &buf);

-        parcel_out.WriteObject(buf);
+        parcel_out.WriteFlattenedObject(buf);
        break;
    }
    case TransactionId::QueueBuffer: {
@@ -117,61 +117,67 @@ private:

 class OutputParcel final {
 public:
-    static constexpr std::size_t DefaultBufferSize = 0x40;
-
-    OutputParcel() : buffer(DefaultBufferSize) {}
-
-    template <typename T>
-    explicit OutputParcel(const T& out_data) : buffer(DefaultBufferSize) {
-        Write(out_data);
-    }
+    OutputParcel() = default;

    template <typename T>
    void Write(const T& val) {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable.");
-
-        if (buffer.size() < write_index + sizeof(T)) {
-            buffer.resize(buffer.size() + sizeof(T) + DefaultBufferSize);
-        }
-
-        std::memcpy(buffer.data() + write_index, &val, sizeof(T));
-        write_index += sizeof(T);
-        write_index = Common::AlignUp(write_index, 4);
+        this->WriteImpl(val, m_data_buffer);
    }

    template <typename T>
-    void WriteObject(const T* ptr) {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable.");
-
+    void WriteFlattenedObject(const T* ptr) {
        if (!ptr) {
-            Write<u32>(0);
+            this->Write<u32>(0);
            return;
        }

-        Write<u32>(1);
-        Write<s64>(sizeof(T));
-        Write(*ptr);
+        this->Write<u32>(1);
+        this->Write<s64>(sizeof(T));
+        this->Write(*ptr);
    }

    template <typename T>
-    void WriteObject(const std::shared_ptr<T> ptr) {
-        WriteObject(ptr.get());
+    void WriteFlattenedObject(const std::shared_ptr<T> ptr) {
+        this->WriteFlattenedObject(ptr.get());
+    }
+
+    template <typename T>
+    void WriteInterface(const T& val) {
+        this->WriteImpl(val, m_data_buffer);
+        this->WriteImpl(0U, m_object_buffer);
    }

    std::vector<u8> Serialize() const {
-        ParcelHeader header{};
-        header.data_size = static_cast<u32>(write_index - sizeof(ParcelHeader));
-        header.data_offset = sizeof(ParcelHeader);
-        header.objects_size = 4;
-        header.objects_offset = static_cast<u32>(sizeof(ParcelHeader) + header.data_size);
-        std::memcpy(buffer.data(), &header, sizeof(ParcelHeader));
+        std::vector<u8> output_buffer(sizeof(ParcelHeader) + m_data_buffer.size() +
+                                      m_object_buffer.size());

-        return buffer;
+        ParcelHeader header{};
+        header.data_size = static_cast<u32>(m_data_buffer.size());
+        header.data_offset = sizeof(ParcelHeader);
+        header.objects_size = static_cast<u32>(m_object_buffer.size());
+        header.objects_offset = header.data_offset + header.data_size;
+
+        std::memcpy(output_buffer.data(), &header, sizeof(header));
+        std::ranges::copy(m_data_buffer, output_buffer.data() + header.data_offset);
+        std::ranges::copy(m_object_buffer, output_buffer.data() + header.objects_offset);
+
+        return output_buffer;
    }

 private:
-    mutable std::vector<u8> buffer;
-    std::size_t write_index = sizeof(ParcelHeader);
+    template <typename T>
+        requires(std::is_trivially_copyable_v<T>)
+    void WriteImpl(const T& val, std::vector<u8>& buffer) {
+        const size_t aligned_size = Common::AlignUp(sizeof(T), 4);
+        const size_t old_size = buffer.size();
+        buffer.resize(old_size + aligned_size);
+
+        std::memcpy(buffer.data() + old_size, &val, sizeof(T));
+    }
+
+private:
+    std::vector<u8> m_data_buffer;
+    std::vector<u8> m_object_buffer;
 };

 } // namespace Service::android
@@ -59,6 +59,18 @@ static_assert(sizeof(SystemClockContext) == 0x20, "SystemClockContext is incorre
 static_assert(std::is_trivially_copyable_v<SystemClockContext>,
              "SystemClockContext must be trivially copyable");

+struct ContinuousAdjustmentTimePoint {
+    s64 measurement_offset;
+    s64 diff_scale;
+    u32 shift_amount;
+    s64 lower;
+    s64 upper;
+    Common::UUID clock_source_id;
+};
+static_assert(sizeof(ContinuousAdjustmentTimePoint) == 0x38);
+static_assert(std::is_trivially_copyable_v<ContinuousAdjustmentTimePoint>,
+              "ContinuousAdjustmentTimePoint must be trivially copyable");
+
 /// https://switchbrew.org/wiki/Glue_services#TimeSpanType
 struct TimeSpanType {
    s64 nanoseconds{};
@@ -30,6 +30,25 @@ void SharedMemory::SetupStandardSteadyClock(const Common::UUID& clock_source_id,
 }

 void SharedMemory::UpdateLocalSystemClockContext(const Clock::SystemClockContext& context) {
+    // lower and upper are related to the measurement point for the steady time point,
+    // and compare equal on boot
+    const s64 time_point_ns = context.steady_time_point.time_point * 1'000'000'000LL;
+
+    // This adjusts for some sort of time skew
+    // Both 0 on boot
+    const s64 diff_scale = 0;
+    const u32 shift_amount = 0;
+
+    const Clock::ContinuousAdjustmentTimePoint adjustment{
+        .measurement_offset = system.CoreTiming().GetGlobalTimeNs().count(),
+        .diff_scale = diff_scale,
+        .shift_amount = shift_amount,
+        .lower = time_point_ns,
+        .upper = time_point_ns,
+        .clock_source_id = context.steady_time_point.clock_source_id,
+    };
+
+    StoreToLockFreeAtomicType(&GetFormat()->continuous_adjustment_timepoint, adjustment);
    StoreToLockFreeAtomicType(&GetFormat()->standard_local_system_clock_context, context);
 }

@@ -65,14 +65,15 @@ public:
        LockFreeAtomicType<Clock::SystemClockContext> standard_local_system_clock_context;
        LockFreeAtomicType<Clock::SystemClockContext> standard_network_system_clock_context;
        LockFreeAtomicType<bool> is_standard_user_system_clock_automatic_correction_enabled;
-        u32 format_version;
+        LockFreeAtomicType<Clock::ContinuousAdjustmentTimePoint> continuous_adjustment_timepoint;
    };
    static_assert(offsetof(Format, standard_steady_clock_timepoint) == 0x0);
    static_assert(offsetof(Format, standard_local_system_clock_context) == 0x38);
    static_assert(offsetof(Format, standard_network_system_clock_context) == 0x80);
    static_assert(offsetof(Format, is_standard_user_system_clock_automatic_correction_enabled) ==
                  0xc8);
-    static_assert(sizeof(Format) == 0xd8, "Format is an invalid size");
+    static_assert(offsetof(Format, continuous_adjustment_timepoint) == 0xd0);
+    static_assert(sizeof(Format) == 0x148, "Format is an invalid size");

    void SetupStandardSteadyClock(const Common::UUID& clock_source_id,
                                  Clock::TimeSpanType current_time_point);
@@ -64,8 +64,8 @@ public:
 private:
    const u32 magic = 2;
    const u32 process_id = 1;
-    const u32 id;
-    INSERT_PADDING_WORDS(3);
+    const u64 id;
+    INSERT_PADDING_WORDS(2);
    std::array<u8, 8> dispdrv = {'d', 'i', 's', 'p', 'd', 'r', 'v', '\0'};
    INSERT_PADDING_WORDS(2);
 };
@@ -608,7 +608,9 @@ private:
            return;
        }

-        const auto parcel = android::OutputParcel{NativeWindow{*buffer_queue_id}};
+        android::OutputParcel parcel;
+        parcel.WriteInterface(NativeWindow{*buffer_queue_id});
+
        const auto buffer_size = ctx.WriteBuffer(parcel.Serialize());

        IPC::ResponseBuilder rb{ctx, 4};
@@ -654,7 +656,9 @@ private:
            return;
        }

-        const auto parcel = android::OutputParcel{NativeWindow{*buffer_queue_id}};
+        android::OutputParcel parcel;
+        parcel.WriteInterface(NativeWindow{*buffer_queue_id});
+
        const auto buffer_size = ctx.WriteBuffer(parcel.Serialize());

        IPC::ResponseBuilder rb{ctx, 6};
@@ -195,8 +195,8 @@ void Joycons::RegisterNewDevice(SDL_hid_device_info* device_info) {
                OnMotionUpdate(port, type, id, value);
            }},
            .on_ring_data = {[this](f32 ring_data) { OnRingConUpdate(ring_data); }},
-            .on_amiibo_data = {[this, port](const std::vector<u8>& amiibo_data) {
-                OnAmiiboUpdate(port, amiibo_data);
+            .on_amiibo_data = {[this, port, type](const std::vector<u8>& amiibo_data) {
+                OnAmiiboUpdate(port, type, amiibo_data);
            }},
            .on_camera_data = {[this, port](const std::vector<u8>& camera_data,
                                            Joycon::IrsResolution format) {
@@ -291,9 +291,13 @@ Common::Input::NfcState Joycons::SupportsNfc(const PadIdentifier& identifier_) c
    return Common::Input::NfcState::Success;
 };

-Common::Input::NfcState Joycons::WriteNfcData(const PadIdentifier& identifier_,
+Common::Input::NfcState Joycons::WriteNfcData(const PadIdentifier& identifier,
                                              const std::vector<u8>& data) {
-    return Common::Input::NfcState::NotSupported;
+    auto handle = GetHandle(identifier);
+    if (handle->WriteNfcData(data) != Joycon::DriverResult::Success) {
+        return Common::Input::NfcState::WriteFailed;
+    }
+    return Common::Input::NfcState::Success;
 };

 Common::Input::DriverResult Joycons::SetPollingMode(const PadIdentifier& identifier,
@@ -398,8 +402,9 @@ void Joycons::OnRingConUpdate(f32 ring_data) {
    SetAxis(identifier, 100, ring_data);
 }

-void Joycons::OnAmiiboUpdate(std::size_t port, const std::vector<u8>& amiibo_data) {
-    const auto identifier = GetIdentifier(port, Joycon::ControllerType::Right);
+void Joycons::OnAmiiboUpdate(std::size_t port, Joycon::ControllerType type,
+                             const std::vector<u8>& amiibo_data) {
+    const auto identifier = GetIdentifier(port, type);
    const auto nfc_state = amiibo_data.empty() ? Common::Input::NfcState::AmiiboRemoved
                                               : Common::Input::NfcState::NewAmiibo;
    SetNfc(identifier, {nfc_state, amiibo_data});
@@ -81,7 +81,8 @@ private:
    void OnMotionUpdate(std::size_t port, Joycon::ControllerType type, int id,
                        const Joycon::MotionData& value);
    void OnRingConUpdate(f32 ring_data);
-    void OnAmiiboUpdate(std::size_t port, const std::vector<u8>& amiibo_data);
+    void OnAmiiboUpdate(std::size_t port, Joycon::ControllerType type,
+                        const std::vector<u8>& amiibo_data);
    void OnCameraUpdate(std::size_t port, const std::vector<u8>& camera_data,
                        Joycon::IrsResolution format);

@@ -109,14 +109,37 @@ public:
    }

    bool RumblePlay(const Common::Input::VibrationStatus vibration) {
-        constexpr u32 rumble_max_duration_ms = 1000;
+        constexpr u32 rumble_max_duration_ms = 2000;
+        constexpr f32 low_start_sensitivity_limit = 140.0;
+        constexpr f32 low_width_sensitivity_limit = 400.0;
+        constexpr f32 high_start_sensitivity_limit = 200.0;
+        constexpr f32 high_width_sensitivity_limit = 700.0;
+        // Try to provide some feeling of the frequency by reducing the amplitude depending on it.
+        f32 low_frequency_scale = 1.0;
+        if (vibration.low_frequency > low_start_sensitivity_limit) {
+            low_frequency_scale =
+                std::max(1.0f - (vibration.low_frequency - low_start_sensitivity_limit) /
+                                    low_width_sensitivity_limit,
+                         0.3f);
+        }
+        f32 low_amplitude = vibration.low_amplitude * low_frequency_scale;
+
+        f32 high_frequency_scale = 1.0;
+        if (vibration.high_frequency > high_start_sensitivity_limit) {
+            high_frequency_scale =
+                std::max(1.0f - (vibration.high_frequency - high_start_sensitivity_limit) /
+                                    high_width_sensitivity_limit,
+                         0.3f);
+        }
+        f32 high_amplitude = vibration.high_amplitude * high_frequency_scale;
+
        if (sdl_controller) {
-            return SDL_GameControllerRumble(
-                       sdl_controller.get(), static_cast<u16>(vibration.low_amplitude),
-                       static_cast<u16>(vibration.high_amplitude), rumble_max_duration_ms) != -1;
+            return SDL_GameControllerRumble(sdl_controller.get(), static_cast<u16>(low_amplitude),
+                                            static_cast<u16>(high_amplitude),
+                                            rumble_max_duration_ms) != -1;
        } else if (sdl_joystick) {
-            return SDL_JoystickRumble(sdl_joystick.get(), static_cast<u16>(vibration.low_amplitude),
-                                      static_cast<u16>(vibration.high_amplitude),
+            return SDL_JoystickRumble(sdl_joystick.get(), static_cast<u16>(low_amplitude),
+                                      static_cast<u16>(high_amplitude),
                                      rumble_max_duration_ms) != -1;
        }

@@ -492,6 +492,26 @@ DriverResult JoyconDriver::SetRingConMode() {
    return result;
 }

+DriverResult JoyconDriver::WriteNfcData(std::span<const u8> data) {
+    std::scoped_lock lock{mutex};
+    disable_input_thread = true;
+
+    if (!supported_features.nfc) {
+        return DriverResult::NotSupported;
+    }
+    if (!nfc_protocol->IsEnabled()) {
+        return DriverResult::Disabled;
+    }
+    if (!amiibo_detected) {
+        return DriverResult::ErrorWritingData;
+    }
+
+    const auto result = nfc_protocol->WriteAmiibo(data);
+
+    disable_input_thread = false;
+    return result;
+}
+
 bool JoyconDriver::IsConnected() const {
    std::scoped_lock lock{mutex};
    return is_connected.load();
@@ -49,6 +49,7 @@ public:
    DriverResult SetIrMode();
    DriverResult SetNfcMode();
    DriverResult SetRingConMode();
+    DriverResult WriteNfcData(std::span<const u8> data);

    void SetCallbacks(const JoyconCallbacks& callbacks);

@@ -265,7 +265,7 @@ DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, MCUSubCom

 DriverResult JoyconCommonProtocol::WaitSetMCUMode(ReportMode report_mode, MCUMode mode) {
    MCUCommandResponse output{};
-    constexpr std::size_t MaxTries{8};
+    constexpr std::size_t MaxTries{16};
    std::size_t tries{};

    do {
@@ -23,6 +23,7 @@ constexpr std::array<u8, 8> DefaultVibrationBuffer{0x0, 0x1, 0x40, 0x40, 0x0, 0x

 using MacAddress = std::array<u8, 6>;
 using SerialNumber = std::array<u8, 15>;
+using TagUUID = std::array<u8, 7>;

 enum class ControllerType : u8 {
    None = 0x00,
@@ -276,12 +277,13 @@ enum class MCUPacketFlag : u8 {
    LastCommandPacket = 0x08,
 };

-enum class NFCReadCommand : u8 {
+enum class NFCCommand : u8 {
    CancelAll = 0x00,
    StartPolling = 0x01,
    StopPolling = 0x02,
    StartWaitingRecieve = 0x04,
-    Ntag = 0x06,
+    ReadNtag = 0x06,
+    WriteNtag = 0x08,
    Mifare = 0x0F,
 };

@@ -292,14 +294,19 @@ enum class NFCTagType : u8 {

 enum class NFCPages {
    Block0 = 0,
+    Block3 = 3,
    Block45 = 45,
    Block135 = 135,
    Block231 = 231,
 };

 enum class NFCStatus : u8 {
+    Ready = 0x00,
+    Polling = 0x01,
    LastPackage = 0x04,
+    WriteDone = 0x05,
    TagLost = 0x07,
+    WriteReady = 0x09,
 };

 enum class IrsMode : u8 {
@@ -394,6 +401,7 @@ enum class DriverResult {
    InvalidHandle,
    NotSupported,
    Disabled,
+    Delayed,
    Unknown,
 };

@@ -558,13 +566,32 @@ static_assert(sizeof(NFCReadBlockCommand) == 0x9, "NFCReadBlockCommand is an inv
 struct NFCReadCommandData {
    u8 unknown;
    u8 uuid_length;
-    u8 unknown_2;
-    std::array<u8, 6> uid;
+    TagUUID uid;
    NFCTagType tag_type;
    NFCReadBlockCommand read_block;
 };
 static_assert(sizeof(NFCReadCommandData) == 0x13, "NFCReadCommandData is an invalid size");

+#pragma pack(push, 1)
+struct NFCWriteCommandData {
+    u8 unknown;
+    u8 uuid_length;
+    TagUUID uid;
+    NFCTagType tag_type;
+    u8 unknown2;
+    u8 unknown3;
+    u8 unknown4;
+    u8 unknown5;
+    u8 unknown6;
+    u8 unknown7;
+    u8 unknown8;
+    u8 magic;
+    u16_be write_count;
+    u8 amiibo_version;
+};
+static_assert(sizeof(NFCWriteCommandData) == 0x15, "NFCWriteCommandData is an invalid size");
+#pragma pack(pop)
+
 struct NFCPollingCommandData {
    u8 enable_mifare;
    u8 unknown_1;
@@ -575,9 +602,9 @@ struct NFCPollingCommandData {
 static_assert(sizeof(NFCPollingCommandData) == 0x05, "NFCPollingCommandData is an invalid size");

 struct NFCRequestState {
-    NFCReadCommand command_argument;
+    NFCCommand command_argument;
+    u8 block_id;
    u8 packet_id;
-    INSERT_PADDING_BYTES(0x1);
    MCUPacketFlag packet_flag;
    u8 data_length;
    union {
@@ -590,6 +617,18 @@ struct NFCRequestState {
 };
 static_assert(sizeof(NFCRequestState) == 0x26, "NFCRequestState is an invalid size");

+struct NFCDataChunk {
+    u8 nfc_page;
+    u8 data_size;
+    std::array<u8, 0xFF> data;
+};
+
+struct NFCWritePackage {
+    NFCWriteCommandData command_data;
+    u8 number_of_chunks;
+    std::array<NFCDataChunk, 4> data_chunks;
+};
+
 struct IrsConfigure {
    MCUCommand command;
    MCUSubCommand sub_command;
@@ -34,6 +34,12 @@ DriverResult NfcProtocol::EnableNfc() {

        result = ConfigureMCU(config);
    }
+    if (result == DriverResult::Success) {
+        result = WaitSetMCUMode(ReportMode::NFC_IR_MODE_60HZ, MCUMode::NFC);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }

    return result;
 }
@@ -56,13 +62,20 @@ DriverResult NfcProtocol::StartNFCPollingMode() {
    LOG_DEBUG(Input, "Start NFC pooling Mode");
    ScopedSetBlocking sb(this);
    DriverResult result{DriverResult::Success};
-    TagFoundData tag_data{};

    if (result == DriverResult::Success) {
-        result = WaitSetMCUMode(ReportMode::NFC_IR_MODE_60HZ, MCUMode::NFC);
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
    }
    if (result == DriverResult::Success) {
-        result = WaitUntilNfcIsReady();
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStartPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Polling);
    }
    if (result == DriverResult::Success) {
        is_enabled = true;
@@ -72,49 +85,99 @@ DriverResult NfcProtocol::StartNFCPollingMode() {
 }

 DriverResult NfcProtocol::ScanAmiibo(std::vector<u8>& data) {
-    LOG_DEBUG(Input, "Start NFC pooling Mode");
+    if (update_counter++ < AMIIBO_UPDATE_DELAY) {
+        return DriverResult::Delayed;
+    }
+    update_counter = 0;
+
+    LOG_DEBUG(Input, "Scan for amiibos");
    ScopedSetBlocking sb(this);
    DriverResult result{DriverResult::Success};
    TagFoundData tag_data{};

    if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
-    }
-    if (result == DriverResult::Success) {
-        result = ReadTag(tag_data);
-    }
-    if (result == DriverResult::Success) {
-        result = WaitUntilNfcIsReady();
-    }
-    if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
+        result = IsTagInRange(tag_data);
    }
+
    if (result == DriverResult::Success) {
+        std::string uuid_string;
+        for (auto& content : tag_data.uuid) {
+            uuid_string += fmt::format(" {:02x}", content);
+        }
+        LOG_INFO(Input, "Tag detected, type={}, uuid={}", tag_data.type, uuid_string);
        result = GetAmiiboData(data);
    }

    return result;
 }

+DriverResult NfcProtocol::WriteAmiibo(std::span<const u8> data) {
+    LOG_DEBUG(Input, "Write amiibo");
+    ScopedSetBlocking sb(this);
+    DriverResult result{DriverResult::Success};
+    TagUUID tag_uuid = GetTagUUID(data);
+    TagFoundData tag_data{};
+
+    if (result == DriverResult::Success) {
+        result = IsTagInRange(tag_data, 7);
+    }
+    if (result == DriverResult::Success) {
+        if (tag_data.uuid != tag_uuid) {
+            result = DriverResult::InvalidParameters;
+        }
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStartPollingRequest(output, true);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::WriteReady);
+    }
+    if (result == DriverResult::Success) {
+        result = WriteAmiiboData(tag_uuid, data);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::WriteDone);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
+    }
+
+    return result;
+}
+
 bool NfcProtocol::HasAmiibo() {
+    if (update_counter++ < AMIIBO_UPDATE_DELAY) {
+        return true;
+    }
+    update_counter = 0;
+
    ScopedSetBlocking sb(this);
    DriverResult result{DriverResult::Success};
    TagFoundData tag_data{};

    if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
+        result = IsTagInRange(tag_data, 7);
    }

    return result == DriverResult::Success;
 }

-DriverResult NfcProtocol::WaitUntilNfcIsReady() {
+DriverResult NfcProtocol::WaitUntilNfcIs(NFCStatus status) {
    constexpr std::size_t timeout_limit = 10;
    MCUCommandResponse output{};
    std::size_t tries = 0;

    do {
-        auto result = SendStartWaitingRecieveRequest(output);
+        auto result = SendNextPackageRequest(output, {});

        if (result != DriverResult::Success) {
            return result;
@@ -124,19 +187,17 @@ DriverResult NfcProtocol::WaitUntilNfcIsReady() {
        }
    } while (output.mcu_report != MCUReport::NFCState ||
             (output.mcu_data[1] << 8) + output.mcu_data[0] != 0x0500 ||
-             output.mcu_data[5] != 0x31 || output.mcu_data[6] != 0x00);
+             output.mcu_data[5] != 0x31 || output.mcu_data[6] != static_cast<u8>(status));

    return DriverResult::Success;
 }

-DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
-    LOG_DEBUG(Input, "Start Polling for tag");
-    constexpr std::size_t timeout_limit = 7;
+DriverResult NfcProtocol::IsTagInRange(TagFoundData& data, std::size_t timeout_limit) {
    MCUCommandResponse output{};
    std::size_t tries = 0;

    do {
-        const auto result = SendStartPollingRequest(output);
+        const auto result = SendNextPackageRequest(output, {});
        if (result != DriverResult::Success) {
            return result;
        }
@@ -145,94 +206,31 @@ DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
        }
    } while (output.mcu_report != MCUReport::NFCState ||
             (output.mcu_data[1] << 8) + output.mcu_data[0] != 0x0500 ||
-             output.mcu_data[6] != 0x09);
+             (output.mcu_data[6] != 0x09 && output.mcu_data[6] != 0x04));

    data.type = output.mcu_data[12];
-    data.uuid.resize(output.mcu_data[14]);
+    data.uuid_size = std::min(output.mcu_data[14], static_cast<u8>(sizeof(TagUUID)));
    memcpy(data.uuid.data(), output.mcu_data.data() + 15, data.uuid.size());

    return DriverResult::Success;
 }

-DriverResult NfcProtocol::ReadTag(const TagFoundData& data) {
-    constexpr std::size_t timeout_limit = 10;
-    MCUCommandResponse output{};
-    std::size_t tries = 0;
-
-    std::string uuid_string;
-    for (auto& content : data.uuid) {
-        uuid_string += fmt::format(" {:02x}", content);
-    }
-
-    LOG_INFO(Input, "Tag detected, type={}, uuid={}", data.type, uuid_string);
-
-    tries = 0;
-    NFCPages ntag_pages = NFCPages::Block0;
-    // Read Tag data
-    while (true) {
-        auto result = SendReadAmiiboRequest(output, ntag_pages);
-        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
-
-        if (result != DriverResult::Success) {
-            return result;
-        }
-
-        if ((output.mcu_report == MCUReport::NFCReadData ||
-             output.mcu_report == MCUReport::NFCState) &&
-            nfc_status == NFCStatus::TagLost) {
-            return DriverResult::ErrorReadingData;
-        }
-
-        if (output.mcu_report == MCUReport::NFCReadData && output.mcu_data[1] == 0x07 &&
-            output.mcu_data[2] == 0x01) {
-            if (data.type != 2) {
-                continue;
-            }
-            switch (output.mcu_data[24]) {
-            case 0:
-                ntag_pages = NFCPages::Block135;
-                break;
-            case 3:
-                ntag_pages = NFCPages::Block45;
-                break;
-            case 4:
-                ntag_pages = NFCPages::Block231;
-                break;
-            default:
-                return DriverResult::ErrorReadingData;
-            }
-            continue;
-        }
-
-        if (output.mcu_report == MCUReport::NFCState && nfc_status == NFCStatus::LastPackage) {
-            // finished
-            SendStopPollingRequest(output);
-            return DriverResult::Success;
-        }
-
-        // Ignore other state reports
-        if (output.mcu_report == MCUReport::NFCState) {
-            continue;
-        }
-
-        if (tries++ > timeout_limit) {
-            return DriverResult::Timeout;
-        }
-    }
-
-    return DriverResult::Success;
-}
-
 DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
-    constexpr std::size_t timeout_limit = 10;
+    constexpr std::size_t timeout_limit = 60;
    MCUCommandResponse output{};
    std::size_t tries = 0;

-    NFCPages ntag_pages = NFCPages::Block135;
+    u8 package_index = 0;
    std::size_t ntag_buffer_pos = 0;
+    auto result = SendReadAmiiboRequest(output, NFCPages::Block135);
+
+    if (result != DriverResult::Success) {
+        return result;
+    }
+
    // Read Tag data
-    while (true) {
-        auto result = SendReadAmiiboRequest(output, ntag_pages);
+    while (tries++ < timeout_limit) {
+        result = SendNextPackageRequest(output, package_index);
        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);

        if (result != DriverResult::Success) {
@@ -255,6 +253,7 @@ DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
                memcpy(ntag_data.data() + ntag_buffer_pos, output.mcu_data.data() + 6,
                       payload_size);
            }
+            package_index++;
            continue;
        }

@@ -262,31 +261,99 @@ DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
            LOG_INFO(Input, "Finished reading amiibo");
            return DriverResult::Success;
        }
+    }

-        // Ignore other state reports
-        if (output.mcu_report == MCUReport::NFCState) {
+    return DriverResult::Timeout;
+}
+
+DriverResult NfcProtocol::WriteAmiiboData(const TagUUID& tag_uuid, std::span<const u8> data) {
+    constexpr std::size_t timeout_limit = 60;
+    const auto nfc_data = MakeAmiiboWritePackage(tag_uuid, data);
+    const std::vector<u8> nfc_buffer_data = SerializeWritePackage(nfc_data);
+    std::span<const u8> buffer(nfc_buffer_data);
+    MCUCommandResponse output{};
+    u8 block_id = 1;
+    u8 package_index = 0;
+    std::size_t tries = 0;
+    std::size_t current_position = 0;
+
+    LOG_INFO(Input, "Writing amiibo data");
+
+    auto result = SendWriteAmiiboRequest(output, tag_uuid);
+
+    if (result != DriverResult::Success) {
+        return result;
+    }
+
+    // Read Tag data but ignore the actual sent data
+    while (tries++ < timeout_limit) {
+        result = SendNextPackageRequest(output, package_index);
+        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
+
+        if (result != DriverResult::Success) {
+            return result;
+        }
+
+        if ((output.mcu_report == MCUReport::NFCReadData ||
+             output.mcu_report == MCUReport::NFCState) &&
+            nfc_status == NFCStatus::TagLost) {
+            return DriverResult::ErrorReadingData;
+        }
+
+        if (output.mcu_report == MCUReport::NFCReadData && output.mcu_data[1] == 0x07) {
+            package_index++;
            continue;
        }

-        if (tries++ > timeout_limit) {
-            return DriverResult::Timeout;
+        if (output.mcu_report == MCUReport::NFCState && nfc_status == NFCStatus::LastPackage) {
+            LOG_INFO(Input, "Finished reading amiibo");
+            break;
        }
    }

-    return DriverResult::Success;
+    // Send Data. Nfc buffer size is 31, Send the data in smaller packages
+    while (current_position < buffer.size() && tries++ < timeout_limit) {
+        const std::size_t next_position =
+            std::min(current_position + sizeof(NFCRequestState::raw_data), buffer.size());
+        const std::size_t block_size = next_position - current_position;
+        const bool is_last_packet = block_size < sizeof(NFCRequestState::raw_data);
+
+        SendWriteDataAmiiboRequest(output, block_id, is_last_packet,
+                                   buffer.subspan(current_position, block_size));
+
+        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
+
+        if ((output.mcu_report == MCUReport::NFCReadData ||
+             output.mcu_report == MCUReport::NFCState) &&
+            nfc_status == NFCStatus::TagLost) {
+            return DriverResult::ErrorReadingData;
+        }
+
+        // Increase position when data is confirmed by the joycon
+        if (output.mcu_report == MCUReport::NFCState &&
+            (output.mcu_data[1] << 8) + output.mcu_data[0] == 0x0500 &&
+            output.mcu_data[3] == block_id) {
+            block_id++;
+            current_position = next_position;
+        }
+    }
+
+    return result;
 }

-DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
+DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output,
+                                                  bool is_second_attempt) {
    NFCRequestState request{
-        .command_argument = NFCReadCommand::StartPolling,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StartPolling,
+        .block_id = {},
+        .packet_id = {},
        .packet_flag = MCUPacketFlag::LastCommandPacket,
        .data_length = sizeof(NFCPollingCommandData),
        .nfc_polling =
            {
-                .enable_mifare = 0x01,
-                .unknown_1 = 0x00,
-                .unknown_2 = 0x00,
+                .enable_mifare = 0x00,
+                .unknown_1 = static_cast<u8>(is_second_attempt ? 0xe8 : 0x00),
+                .unknown_2 = static_cast<u8>(is_second_attempt ? 0x03 : 0x00),
                .unknown_3 = 0x2c,
                .unknown_4 = 0x01,
            },
@@ -302,10 +369,11 @@ DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {

 DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
    NFCRequestState request{
-        .command_argument = NFCReadCommand::StopPolling,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StopPolling,
+        .block_id = {},
+        .packet_id = {},
        .packet_flag = MCUPacketFlag::LastCommandPacket,
-        .data_length = 0,
+        .data_length = {},
        .raw_data = {},
        .crc = {},
    };
@@ -317,12 +385,13 @@ DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
                       output);
 }

-DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& output) {
+DriverResult NfcProtocol::SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id) {
    NFCRequestState request{
-        .command_argument = NFCReadCommand::StartWaitingRecieve,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StartWaitingRecieve,
+        .block_id = {},
+        .packet_id = packet_id,
        .packet_flag = MCUPacketFlag::LastCommandPacket,
-        .data_length = 0,
+        .data_length = {},
        .raw_data = {},
        .crc = {},
    };
@@ -336,17 +405,17 @@ DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& out

 DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages) {
    NFCRequestState request{
-        .command_argument = NFCReadCommand::Ntag,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::ReadNtag,
+        .block_id = {},
+        .packet_id = {},
        .packet_flag = MCUPacketFlag::LastCommandPacket,
        .data_length = sizeof(NFCReadCommandData),
        .nfc_read =
            {
                .unknown = 0xd0,
-                .uuid_length = 0x07,
-                .unknown_2 = 0x00,
+                .uuid_length = sizeof(NFCReadCommandData::uid),
                .uid = {},
-                .tag_type = NFCTagType::AllTags,
+                .tag_type = NFCTagType::Ntag215,
                .read_block = GetReadBlockCommand(ntag_pages),
            },
        .crc = {},
@@ -359,12 +428,135 @@ DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCP
                       output);
 }

+DriverResult NfcProtocol::SendWriteAmiiboRequest(MCUCommandResponse& output,
+                                                 const TagUUID& tag_uuid) {
+    NFCRequestState request{
+        .command_argument = NFCCommand::ReadNtag,
+        .block_id = {},
+        .packet_id = {},
+        .packet_flag = MCUPacketFlag::LastCommandPacket,
+        .data_length = sizeof(NFCReadCommandData),
+        .nfc_read =
+            {
+                .unknown = 0xd0,
+                .uuid_length = sizeof(NFCReadCommandData::uid),
+                .uid = tag_uuid,
+                .tag_type = NFCTagType::Ntag215,
+                .read_block = GetReadBlockCommand(NFCPages::Block3),
+            },
+        .crc = {},
+    };
+
+    std::array<u8, sizeof(NFCRequestState)> request_data{};
+    memcpy(request_data.data(), &request, sizeof(NFCRequestState));
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
+}
+
+DriverResult NfcProtocol::SendWriteDataAmiiboRequest(MCUCommandResponse& output, u8 block_id,
+                                                     bool is_last_packet,
+                                                     std::span<const u8> data) {
+    const auto data_size = std::min(data.size(), sizeof(NFCRequestState::raw_data));
+    NFCRequestState request{
+        .command_argument = NFCCommand::WriteNtag,
+        .block_id = block_id,
+        .packet_id = {},
+        .packet_flag =
+            is_last_packet ? MCUPacketFlag::LastCommandPacket : MCUPacketFlag::MorePacketsRemaining,
+        .data_length = static_cast<u8>(data_size),
+        .raw_data = {},
+        .crc = {},
+    };
+    memcpy(request.raw_data.data(), data.data(), data_size);
+
+    std::array<u8, sizeof(NFCRequestState)> request_data{};
+    memcpy(request_data.data(), &request, sizeof(NFCRequestState));
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
+}
+
+std::vector<u8> NfcProtocol::SerializeWritePackage(const NFCWritePackage& package) const {
+    const std::size_t header_size =
+        sizeof(NFCWriteCommandData) + sizeof(NFCWritePackage::number_of_chunks);
+    std::vector<u8> serialized_data(header_size);
+    std::size_t start_index = 0;
+
+    memcpy(serialized_data.data(), &package, header_size);
+    start_index += header_size;
+
+    for (const auto& data_chunk : package.data_chunks) {
+        const std::size_t chunk_size =
+            sizeof(NFCDataChunk::nfc_page) + sizeof(NFCDataChunk::data_size) + data_chunk.data_size;
+
+        serialized_data.resize(start_index + chunk_size);
+        memcpy(serialized_data.data() + start_index, &data_chunk, chunk_size);
+        start_index += chunk_size;
+    }
+
+    return serialized_data;
+}
+
+NFCWritePackage NfcProtocol::MakeAmiiboWritePackage(const TagUUID& tag_uuid,
+                                                    std::span<const u8> data) const {
+    return {
+        .command_data{
+            .unknown = 0xd0,
+            .uuid_length = sizeof(NFCReadCommandData::uid),
+            .uid = tag_uuid,
+            .tag_type = NFCTagType::Ntag215,
+            .unknown2 = 0x00,
+            .unknown3 = 0x01,
+            .unknown4 = 0x04,
+            .unknown5 = 0xff,
+            .unknown6 = 0xff,
+            .unknown7 = 0xff,
+            .unknown8 = 0xff,
+            .magic = data[16],
+            .write_count = static_cast<u16>((data[17] << 8) + data[18]),
+            .amiibo_version = data[19],
+        },
+        .number_of_chunks = 3,
+        .data_chunks =
+            {
+                MakeAmiiboChunk(0x05, 0x20, data),
+                MakeAmiiboChunk(0x20, 0xf0, data),
+                MakeAmiiboChunk(0x5c, 0x98, data),
+            },
+    };
+}
+
+NFCDataChunk NfcProtocol::MakeAmiiboChunk(u8 page, u8 size, std::span<const u8> data) const {
+    constexpr u8 NFC_PAGE_SIZE = 4;
+
+    if (static_cast<std::size_t>(page * NFC_PAGE_SIZE) + size >= data.size()) {
+        return {};
+    }
+
+    NFCDataChunk chunk{
+        .nfc_page = page,
+        .data_size = size,
+        .data = {},
+    };
+    std::memcpy(chunk.data.data(), data.data() + (page * NFC_PAGE_SIZE), size);
+    return chunk;
+}
+
 NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {
    switch (pages) {
    case NFCPages::Block0:
        return {
            .block_count = 1,
        };
+    case NFCPages::Block3:
+        return {
+            .block_count = 1,
+            .blocks =
+                {
+                    NFCReadBlock{0x03, 0x03},
+                },
+        };
    case NFCPages::Block45:
        return {
            .block_count = 1,
@@ -399,6 +591,17 @@ NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {
    };
 }

+TagUUID NfcProtocol::GetTagUUID(std::span<const u8> data) const {
+    if (data.size() < 10) {
+        return {};
+    }
+
+    // crc byte 3 is omitted in this operation
+    return {
+        data[0], data[1], data[2], data[4], data[5], data[6], data[7],
+    };
+}
+
 bool NfcProtocol::IsEnabled() const {
    return is_enabled;
 }
@@ -27,35 +27,56 @@ public:

    DriverResult ScanAmiibo(std::vector<u8>& data);

+    DriverResult WriteAmiibo(std::span<const u8> data);
+
    bool HasAmiibo();

    bool IsEnabled() const;

 private:
+    // Number of times the function will be delayed until it outputs valid data
+    static constexpr std::size_t AMIIBO_UPDATE_DELAY = 15;
+
    struct TagFoundData {
        u8 type;
-        std::vector<u8> uuid;
+        u8 uuid_size;
+        TagUUID uuid;
    };

-    DriverResult WaitUntilNfcIsReady();
+    DriverResult WaitUntilNfcIs(NFCStatus status);

-    DriverResult StartPolling(TagFoundData& data);
-
-    DriverResult ReadTag(const TagFoundData& data);
+    DriverResult IsTagInRange(TagFoundData& data, std::size_t timeout_limit = 1);

    DriverResult GetAmiiboData(std::vector<u8>& data);

-    DriverResult SendStartPollingRequest(MCUCommandResponse& output);
+    DriverResult WriteAmiiboData(const TagUUID& tag_uuid, std::span<const u8> data);
+
+    DriverResult SendStartPollingRequest(MCUCommandResponse& output,
+                                         bool is_second_attempt = false);

    DriverResult SendStopPollingRequest(MCUCommandResponse& output);

-    DriverResult SendStartWaitingRecieveRequest(MCUCommandResponse& output);
+    DriverResult SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id);

    DriverResult SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages);

+    DriverResult SendWriteAmiiboRequest(MCUCommandResponse& output, const TagUUID& tag_uuid);
+
+    DriverResult SendWriteDataAmiiboRequest(MCUCommandResponse& output, u8 block_id,
+                                            bool is_last_packet, std::span<const u8> data);
+
+    std::vector<u8> SerializeWritePackage(const NFCWritePackage& package) const;
+
+    NFCWritePackage MakeAmiiboWritePackage(const TagUUID& tag_uuid, std::span<const u8> data) const;
+
+    NFCDataChunk MakeAmiiboChunk(u8 page, u8 size, std::span<const u8> data) const;
+
    NFCReadBlockCommand GetReadBlockCommand(NFCPages pages) const;

+    TagUUID GetTagUUID(std::span<const u8> data) const;
+
    bool is_enabled{};
+    std::size_t update_counter{};
 };

 } // namespace InputCommon::Joycon
@@ -380,13 +380,16 @@ void InputEngine::TriggerOnMotionChange(const PadIdentifier& identifier, int mot
    if (!configuring || !mapping_callback.on_data) {
        return;
    }
+    const auto old_value = GetMotion(identifier, motion);
    bool is_active = false;
-    if (std::abs(value.accel_x) > 1.5f || std::abs(value.accel_y) > 1.5f ||
-        std::abs(value.accel_z) > 1.5f) {
+    if (std::abs(value.accel_x - old_value.accel_x) > 1.5f ||
+        std::abs(value.accel_y - old_value.accel_y) > 1.5f ||
+        std::abs(value.accel_z - old_value.accel_z) > 1.5f) {
        is_active = true;
    }
-    if (std::abs(value.gyro_x) > 0.6f || std::abs(value.gyro_y) > 0.6f ||
-        std::abs(value.gyro_z) > 0.6f) {
+    if (std::abs(value.gyro_x - old_value.gyro_x) > 0.6f ||
+        std::abs(value.gyro_y - old_value.gyro_y) > 0.6f ||
+        std::abs(value.gyro_z - old_value.gyro_z) > 0.6f) {
        is_active = true;
    }
    if (!is_active) {
@@ -339,9 +339,7 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
        if (ctx.profile.support_vertex_instance_id) {
            return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.vertex_id));
        } else {
-            const Id index{ctx.OpLoad(ctx.U32[1], ctx.vertex_index)};
-            const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
-            return ctx.OpBitcast(ctx.F32[1], ctx.OpISub(ctx.U32[1], index, base));
+            return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.vertex_index));
        }
    case IR::Attribute::BaseInstance:
        return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.base_instance));
@@ -386,9 +384,7 @@ Id EmitGetAttributeU32(EmitContext& ctx, IR::Attribute attr, Id) {
        if (ctx.profile.support_vertex_instance_id) {
            return ctx.OpLoad(ctx.U32[1], ctx.vertex_id);
        } else {
-            const Id index{ctx.OpLoad(ctx.U32[1], ctx.vertex_index)};
-            const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
-            return ctx.OpISub(ctx.U32[1], index, base);
+            return ctx.OpLoad(ctx.U32[1], ctx.vertex_index);
        }
    case IR::Attribute::BaseInstance:
        return ctx.OpLoad(ctx.U32[1], ctx.base_instance);
@@ -30,7 +30,7 @@ void SHF(TranslatorVisitor& v, u64 insn, const IR::U32& shift, const IR::U32& hi
    union {
        u64 insn;
        BitField<0, 8, IR::Reg> dest_reg;
-        BitField<0, 8, IR::Reg> lo_bits_reg;
+        BitField<8, 8, IR::Reg> lo_bits_reg;
        BitField<37, 2, MaxShift> max_shift;
        BitField<47, 1, u64> cc;
        BitField<48, 2, u64> x_mode;
@@ -102,12 +102,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
        }
        IR::F32 value{v.ir.CompositeExtract(sample, element)};
        if (element < 2) {
-            IR::U32 casted_value;
-            if (element == 0) {
-                casted_value = v.ir.ConvertFToU(32, value);
-            } else {
-                casted_value = v.ir.ConvertFToS(16, value);
-            }
+            IR::U32 casted_value = v.ir.ConvertFToU(32, value);
            v.X(dest_reg, v.ir.ShiftLeftLogical(casted_value, v.ir.Imm32(8)));
        } else {
            v.F(dest_reg, value);
@@ -246,10 +246,14 @@ add_library(video_core STATIC
    texture_cache/util.h
    textures/astc.h
    textures/astc.cpp
+    textures/bcn.cpp
+    textures/bcn.h
    textures/decoders.cpp
    textures/decoders.h
    textures/texture.cpp
    textures/texture.h
+    textures/workers.cpp
+    textures/workers.h
    transform_feedback.cpp
    transform_feedback.h
    video_core.cpp
@@ -275,7 +279,7 @@ add_library(video_core STATIC
 create_target_directory_groups(video_core)

 target_link_libraries(video_core PUBLIC common core)
-target_link_libraries(video_core PUBLIC glad shader_recompiler)
+target_link_libraries(video_core PUBLIC glad shader_recompiler stb)

 if (YUZU_USE_BUNDLED_FFMPEG AND NOT WIN32)
    add_dependencies(video_core ffmpeg-build)
@@ -2,6 +2,8 @@
 // SPDX-License-Identifier: GPL-3.0-or-later

 #include "common/microprofile.h"
+#include "video_core/buffer_cache/buffer_cache_base.h"
+#include "video_core/control/channel_state_cache.inc"

 namespace VideoCommon {

@@ -9,4 +11,6 @@ MICROPROFILE_DEFINE(GPU_PrepareBuffers, "GPU", "Prepare buffers", MP_RGB(224, 12
 MICROPROFILE_DEFINE(GPU_BindUploadBuffers, "GPU", "Bind and upload buffers", MP_RGB(224, 128, 128));
 MICROPROFILE_DEFINE(GPU_DownloadMemory, "GPU", "Download buffers", MP_RGB(224, 128, 128));

+template class VideoCommon::ChannelSetupCaches<VideoCommon::BufferCacheChannelInfo>;
+
 } // namespace VideoCommon
@@ -30,8 +30,8 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
    }

    const s64 device_memory = static_cast<s64>(runtime.GetDeviceLocalMemory());
-    const s64 min_spacing_expected = device_memory - 1_GiB - 512_MiB;
-    const s64 min_spacing_critical = device_memory - 1_GiB;
+    const s64 min_spacing_expected = device_memory - 1_GiB;
+    const s64 min_spacing_critical = device_memory - 512_MiB;
    const s64 mem_threshold = std::min(device_memory, TARGET_THRESHOLD);
    const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
    const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
@@ -63,18 +63,27 @@ void BufferCache<P>::RunGarbageCollector() {

 template <class P>
 void BufferCache<P>::TickFrame() {
+    // Homebrew console apps don't create or bind any channels, so this will be nullptr.
+    if (!channel_state) {
+        return;
+    }
+
    // Calculate hits and shots and move hit bits to the right
-    const u32 hits = std::reduce(uniform_cache_hits.begin(), uniform_cache_hits.end());
-    const u32 shots = std::reduce(uniform_cache_shots.begin(), uniform_cache_shots.end());
-    std::copy_n(uniform_cache_hits.begin(), uniform_cache_hits.size() - 1,
-                uniform_cache_hits.begin() + 1);
-    std::copy_n(uniform_cache_shots.begin(), uniform_cache_shots.size() - 1,
-                uniform_cache_shots.begin() + 1);
-    uniform_cache_hits[0] = 0;
-    uniform_cache_shots[0] = 0;
+    const u32 hits = std::reduce(channel_state->uniform_cache_hits.begin(),
+                                 channel_state->uniform_cache_hits.end());
+    const u32 shots = std::reduce(channel_state->uniform_cache_shots.begin(),
+                                  channel_state->uniform_cache_shots.end());
+    std::copy_n(channel_state->uniform_cache_hits.begin(),
+                channel_state->uniform_cache_hits.size() - 1,
+                channel_state->uniform_cache_hits.begin() + 1);
+    std::copy_n(channel_state->uniform_cache_shots.begin(),
+                channel_state->uniform_cache_shots.size() - 1,
+                channel_state->uniform_cache_shots.begin() + 1);
+    channel_state->uniform_cache_hits[0] = 0;
+    channel_state->uniform_cache_shots[0] = 0;

    const bool skip_preferred = hits * 256 < shots * 251;
-    uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;
+    channel_state->uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;

    // If we can obtain the memory info, use it instead of the estimate.
    if (runtime.CanReportMemoryUsage()) {
@@ -96,12 +105,12 @@ void BufferCache<P>::TickFrame() {

 template <class P>
 void BufferCache<P>::WriteMemory(VAddr cpu_addr, u64 size) {
-    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
    if (memory_tracker.IsRegionGpuModified(cpu_addr, size)) {
        const IntervalType subtract_interval{cpu_addr, cpu_addr + size};
        ClearDownload(subtract_interval);
        common_ranges.subtract(subtract_interval);
    }
+    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
 }

 template <class P>
@@ -122,41 +131,24 @@ std::optional<VideoCore::RasterizerDownloadArea> BufferCache<P>::GetFlushArea(VA
        area->preemtive = true;
        return area;
    };
+    area->preemtive =
+        !IsRegionGpuModified(cpu_addr_start_aligned, cpu_addr_end_aligned - cpu_addr_start_aligned);
    memory_tracker.MarkRegionAsPreflushable(cpu_addr_start_aligned,
                                            cpu_addr_end_aligned - cpu_addr_start_aligned);
-    area->preemtive = !IsRegionGpuModified(cpu_addr, size);
    return area;
 }

 template <class P>
 void BufferCache<P>::DownloadMemory(VAddr cpu_addr, u64 size) {
-    WaitOnAsyncFlushes(cpu_addr, size);
    ForEachBufferInRange(cpu_addr, size, [&](BufferId, Buffer& buffer) {
        DownloadBufferMemory(buffer, cpu_addr, size);
    });
 }

-template <class P>
-void BufferCache<P>::WaitOnAsyncFlushes(VAddr cpu_addr, u64 size) {
-    bool must_wait = false;
-    ForEachInOverlapCounter(async_downloads, cpu_addr, size,
-                            [&](VAddr, VAddr, int) { must_wait = true; });
-    bool must_release = false;
-    ForEachInRangeSet(pending_ranges, cpu_addr, size, [&](VAddr, VAddr) { must_release = true; });
-    if (must_release) {
-        std::function<void()> tmp([]() {});
-        rasterizer.SignalFence(std::move(tmp));
-    }
-    if (must_wait || must_release) {
-        rasterizer.ReleaseFences();
-    }
-}
-
 template <class P>
 void BufferCache<P>::ClearDownload(IntervalType subtract_interval) {
    RemoveEachInOverlapCounter(async_downloads, subtract_interval, -1024);
    uncommitted_ranges.subtract(subtract_interval);
-    pending_ranges.subtract(subtract_interval);
    for (auto& interval_set : committed_ranges) {
        interval_set.subtract(subtract_interval);
    }
@@ -176,16 +168,15 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
    }

    const IntervalType subtract_interval{*cpu_dest_address, *cpu_dest_address + amount};
-    WaitOnAsyncFlushes(*cpu_src_address, static_cast<u32>(amount));
    ClearDownload(subtract_interval);

    BufferId buffer_a;
    BufferId buffer_b;
    do {
-        has_deleted_buffers = false;
+        channel_state->has_deleted_buffers = false;
        buffer_a = FindBuffer(*cpu_src_address, static_cast<u32>(amount));
        buffer_b = FindBuffer(*cpu_dest_address, static_cast<u32>(amount));
-    } while (has_deleted_buffers);
+    } while (channel_state->has_deleted_buffers);
    auto& src_buffer = slot_buffers[buffer_a];
    auto& dest_buffer = slot_buffers[buffer_b];
    SynchronizeBuffer(src_buffer, *cpu_src_address, static_cast<u32>(amount));
@@ -204,7 +195,6 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
        const IntervalType add_interval{new_base_address, new_base_address + size};
        tmp_intervals.push_back(add_interval);
        uncommitted_ranges.add(add_interval);
-        pending_ranges.add(add_interval);
    };
    ForEachInRangeSet(common_ranges, *cpu_src_address, amount, mirror);
    // This subtraction in this order is important for overlapping copies.
@@ -291,30 +281,30 @@ void BufferCache<P>::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr
        .size = size,
        .buffer_id = BufferId{},
    };
-    uniform_buffers[stage][index] = binding;
+    channel_state->uniform_buffers[stage][index] = binding;
 }

 template <class P>
 void BufferCache<P>::DisableGraphicsUniformBuffer(size_t stage, u32 index) {
-    uniform_buffers[stage][index] = NULL_BINDING;
+    channel_state->uniform_buffers[stage][index] = NULL_BINDING;
 }

 template <class P>
 void BufferCache<P>::UpdateGraphicsBuffers(bool is_indexed) {
    MICROPROFILE_SCOPE(GPU_PrepareBuffers);
    do {
-        has_deleted_buffers = false;
+        channel_state->has_deleted_buffers = false;
        DoUpdateGraphicsBuffers(is_indexed);
-    } while (has_deleted_buffers);
+    } while (channel_state->has_deleted_buffers);
 }

 template <class P>
 void BufferCache<P>::UpdateComputeBuffers() {
    MICROPROFILE_SCOPE(GPU_PrepareBuffers);
    do {
-        has_deleted_buffers = false;
+        channel_state->has_deleted_buffers = false;
        DoUpdateComputeBuffers();
-    } while (has_deleted_buffers);
+    } while (channel_state->has_deleted_buffers);
 }

 template <class P>
@@ -357,98 +347,102 @@ template <class P>
 void BufferCache<P>::SetUniformBuffersState(const std::array<u32, NUM_STAGES>& mask,
                                            const UniformBufferSizes* sizes) {
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-        if (enabled_uniform_buffer_masks != mask) {
+        if (channel_state->enabled_uniform_buffer_masks != mask) {
            if constexpr (IS_OPENGL) {
-                fast_bound_uniform_buffers.fill(0);
+                channel_state->fast_bound_uniform_buffers.fill(0);
            }
-            dirty_uniform_buffers.fill(~u32{0});
-            uniform_buffer_binding_sizes.fill({});
+            channel_state->dirty_uniform_buffers.fill(~u32{0});
+            channel_state->uniform_buffer_binding_sizes.fill({});
        }
    }
-    enabled_uniform_buffer_masks = mask;
-    uniform_buffer_sizes = sizes;
+    channel_state->enabled_uniform_buffer_masks = mask;
+    channel_state->uniform_buffer_sizes = sizes;
 }

 template <class P>
 void BufferCache<P>::SetComputeUniformBufferState(u32 mask,
                                                  const ComputeUniformBufferSizes* sizes) {
-    enabled_compute_uniform_buffer_mask = mask;
-    compute_uniform_buffer_sizes = sizes;
+    channel_state->enabled_compute_uniform_buffer_mask = mask;
+    channel_state->compute_uniform_buffer_sizes = sizes;
 }

 template <class P>
 void BufferCache<P>::UnbindGraphicsStorageBuffers(size_t stage) {
-    enabled_storage_buffers[stage] = 0;
-    written_storage_buffers[stage] = 0;
+    channel_state->enabled_storage_buffers[stage] = 0;
+    channel_state->written_storage_buffers[stage] = 0;
 }

 template <class P>
 void BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index,
                                               u32 cbuf_offset, bool is_written) {
-    enabled_storage_buffers[stage] |= 1U << ssbo_index;
-    written_storage_buffers[stage] |= (is_written ? 1U : 0U) << ssbo_index;
+    channel_state->enabled_storage_buffers[stage] |= 1U << ssbo_index;
+    channel_state->written_storage_buffers[stage] |= (is_written ? 1U : 0U) << ssbo_index;

    const auto& cbufs = maxwell3d->state.shader_stages[stage];
    const GPUVAddr ssbo_addr = cbufs.const_buffers[cbuf_index].address + cbuf_offset;
-    storage_buffers[stage][ssbo_index] = StorageBufferBinding(ssbo_addr, cbuf_index, is_written);
+    channel_state->storage_buffers[stage][ssbo_index] =
+        StorageBufferBinding(ssbo_addr, cbuf_index, is_written);
 }

 template <class P>
 void BufferCache<P>::UnbindGraphicsTextureBuffers(size_t stage) {
-    enabled_texture_buffers[stage] = 0;
-    written_texture_buffers[stage] = 0;
-    image_texture_buffers[stage] = 0;
+    channel_state->enabled_texture_buffers[stage] = 0;
+    channel_state->written_texture_buffers[stage] = 0;
+    channel_state->image_texture_buffers[stage] = 0;
 }

 template <class P>
 void BufferCache<P>::BindGraphicsTextureBuffer(size_t stage, size_t tbo_index, GPUVAddr gpu_addr,
                                               u32 size, PixelFormat format, bool is_written,
                                               bool is_image) {
-    enabled_texture_buffers[stage] |= 1U << tbo_index;
-    written_texture_buffers[stage] |= (is_written ? 1U : 0U) << tbo_index;
+    channel_state->enabled_texture_buffers[stage] |= 1U << tbo_index;
+    channel_state->written_texture_buffers[stage] |= (is_written ? 1U : 0U) << tbo_index;
    if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
-        image_texture_buffers[stage] |= (is_image ? 1U : 0U) << tbo_index;
+        channel_state->image_texture_buffers[stage] |= (is_image ? 1U : 0U) << tbo_index;
    }
-    texture_buffers[stage][tbo_index] = GetTextureBufferBinding(gpu_addr, size, format);
+    channel_state->texture_buffers[stage][tbo_index] =
+        GetTextureBufferBinding(gpu_addr, size, format);
 }

 template <class P>
 void BufferCache<P>::UnbindComputeStorageBuffers() {
-    enabled_compute_storage_buffers = 0;
-    written_compute_storage_buffers = 0;
-    image_compute_texture_buffers = 0;
+    channel_state->enabled_compute_storage_buffers = 0;
+    channel_state->written_compute_storage_buffers = 0;
+    channel_state->image_compute_texture_buffers = 0;
 }

 template <class P>
 void BufferCache<P>::BindComputeStorageBuffer(size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
                                              bool is_written) {
-    enabled_compute_storage_buffers |= 1U << ssbo_index;
-    written_compute_storage_buffers |= (is_written ? 1U : 0U) << ssbo_index;
+    channel_state->enabled_compute_storage_buffers |= 1U << ssbo_index;
+    channel_state->written_compute_storage_buffers |= (is_written ? 1U : 0U) << ssbo_index;

    const auto& launch_desc = kepler_compute->launch_description;
    ASSERT(((launch_desc.const_buffer_enable_mask >> cbuf_index) & 1) != 0);

    const auto& cbufs = launch_desc.const_buffer_config;
    const GPUVAddr ssbo_addr = cbufs[cbuf_index].Address() + cbuf_offset;
-    compute_storage_buffers[ssbo_index] = StorageBufferBinding(ssbo_addr, cbuf_index, is_written);
+    channel_state->compute_storage_buffers[ssbo_index] =
+        StorageBufferBinding(ssbo_addr, cbuf_index, is_written);
 }

 template <class P>
 void BufferCache<P>::UnbindComputeTextureBuffers() {
-    enabled_compute_texture_buffers = 0;
-    written_compute_texture_buffers = 0;
-    image_compute_texture_buffers = 0;
+    channel_state->enabled_compute_texture_buffers = 0;
+    channel_state->written_compute_texture_buffers = 0;
+    channel_state->image_compute_texture_buffers = 0;
 }

 template <class P>
 void BufferCache<P>::BindComputeTextureBuffer(size_t tbo_index, GPUVAddr gpu_addr, u32 size,
                                              PixelFormat format, bool is_written, bool is_image) {
-    enabled_compute_texture_buffers |= 1U << tbo_index;
-    written_compute_texture_buffers |= (is_written ? 1U : 0U) << tbo_index;
+    channel_state->enabled_compute_texture_buffers |= 1U << tbo_index;
+    channel_state->written_compute_texture_buffers |= (is_written ? 1U : 0U) << tbo_index;
    if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
-        image_compute_texture_buffers |= (is_image ? 1U : 0U) << tbo_index;
+        channel_state->image_compute_texture_buffers |= (is_image ? 1U : 0U) << tbo_index;
    }
-    compute_texture_buffers[tbo_index] = GetTextureBufferBinding(gpu_addr, size, format);
+    channel_state->compute_texture_buffers[tbo_index] =
+        GetTextureBufferBinding(gpu_addr, size, format);
 }

 template <class P>
@@ -491,7 +485,6 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
    }
    MICROPROFILE_SCOPE(GPU_DownloadMemory);

-    pending_ranges.clear();
    auto it = committed_ranges.begin();
    while (it != committed_ranges.end()) {
        auto& current_intervals = *it;
@@ -692,10 +685,10 @@ bool BufferCache<P>::IsRegionCpuModified(VAddr addr, size_t size) {

 template <class P>
 void BufferCache<P>::BindHostIndexBuffer() {
-    Buffer& buffer = slot_buffers[index_buffer.buffer_id];
-    TouchBuffer(buffer, index_buffer.buffer_id);
-    const u32 offset = buffer.Offset(index_buffer.cpu_addr);
-    const u32 size = index_buffer.size;
+    Buffer& buffer = slot_buffers[channel_state->index_buffer.buffer_id];
+    TouchBuffer(buffer, channel_state->index_buffer.buffer_id);
+    const u32 offset = buffer.Offset(channel_state->index_buffer.cpu_addr);
+    const u32 size = channel_state->index_buffer.size;
    const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
    if (!draw_state.inline_index_draw_indexes.empty()) [[unlikely]] {
        if constexpr (USE_MEMORY_MAPS) {
@@ -709,7 +702,7 @@ void BufferCache<P>::BindHostIndexBuffer() {
            buffer.ImmediateUpload(0, draw_state.inline_index_draw_indexes);
        }
    } else {
-        SynchronizeBuffer(buffer, index_buffer.cpu_addr, size);
+        SynchronizeBuffer(buffer, channel_state->index_buffer.cpu_addr, size);
    }
    if constexpr (HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT) {
        const u32 new_offset =
@@ -726,7 +719,7 @@ template <class P>
 void BufferCache<P>::BindHostVertexBuffers() {
    auto& flags = maxwell3d->dirty.flags;
    for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
-        const Binding& binding = vertex_buffers[index];
+        const Binding& binding = channel_state->vertex_buffers[index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        TouchBuffer(buffer, binding.buffer_id);
        SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
@@ -749,19 +742,19 @@ void BufferCache<P>::BindHostDrawIndirectBuffers() {
        SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
    };
    if (current_draw_indirect->include_count) {
-        bind_buffer(count_buffer_binding);
+        bind_buffer(channel_state->count_buffer_binding);
    }
-    bind_buffer(indirect_buffer_binding);
+    bind_buffer(channel_state->indirect_buffer_binding);
 }

 template <class P>
 void BufferCache<P>::BindHostGraphicsUniformBuffers(size_t stage) {
    u32 dirty = ~0U;
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-        dirty = std::exchange(dirty_uniform_buffers[stage], 0);
+        dirty = std::exchange(channel_state->dirty_uniform_buffers[stage], 0);
    }
    u32 binding_index = 0;
-    ForEachEnabledBit(enabled_uniform_buffer_masks[stage], [&](u32 index) {
+    ForEachEnabledBit(channel_state->enabled_uniform_buffer_masks[stage], [&](u32 index) {
        const bool needs_bind = ((dirty >> index) & 1) != 0;
        BindHostGraphicsUniformBuffer(stage, index, binding_index, needs_bind);
        if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
@@ -773,13 +766,13 @@ void BufferCache<P>::BindHostGraphicsUniformBuffers(size_t stage) {
 template <class P>
 void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32 binding_index,
                                                   bool needs_bind) {
-    const Binding& binding = uniform_buffers[stage][index];
+    const Binding& binding = channel_state->uniform_buffers[stage][index];
    const VAddr cpu_addr = binding.cpu_addr;
-    const u32 size = std::min(binding.size, (*uniform_buffer_sizes)[stage][index]);
+    const u32 size = std::min(binding.size, (*channel_state->uniform_buffer_sizes)[stage][index]);
    Buffer& buffer = slot_buffers[binding.buffer_id];
    TouchBuffer(buffer, binding.buffer_id);
    const bool use_fast_buffer = binding.buffer_id != NULL_BUFFER_ID &&
-                                 size <= uniform_buffer_skip_cache_size &&
+                                 size <= channel_state->uniform_buffer_skip_cache_size &&
                                 !memory_tracker.IsRegionGpuModified(cpu_addr, size);
    if (use_fast_buffer) {
        if constexpr (IS_OPENGL) {
@@ -787,11 +780,11 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
                // Fast path for Nvidia
                const bool should_fast_bind =
                    !HasFastUniformBufferBound(stage, binding_index) ||
-                    uniform_buffer_binding_sizes[stage][binding_index] != size;
+                    channel_state->uniform_buffer_binding_sizes[stage][binding_index] != size;
                if (should_fast_bind) {
                    // We only have to bind when the currently bound buffer is not the fast version
-                    fast_bound_uniform_buffers[stage] |= 1U << binding_index;
-                    uniform_buffer_binding_sizes[stage][binding_index] = size;
+                    channel_state->fast_bound_uniform_buffers[stage] |= 1U << binding_index;
+                    channel_state->uniform_buffer_binding_sizes[stage][binding_index] = size;
                    runtime.BindFastUniformBuffer(stage, binding_index, size);
                }
                const auto span = ImmediateBufferWithData(cpu_addr, size);
@@ -800,8 +793,8 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
            }
        }
        if constexpr (IS_OPENGL) {
-            fast_bound_uniform_buffers[stage] |= 1U << binding_index;
-            uniform_buffer_binding_sizes[stage][binding_index] = size;
+            channel_state->fast_bound_uniform_buffers[stage] |= 1U << binding_index;
+            channel_state->uniform_buffer_binding_sizes[stage][binding_index] = size;
        }
        // Stream buffer path to avoid stalling on non-Nvidia drivers or Vulkan
        const std::span<u8> span = runtime.BindMappedUniformBuffer(stage, binding_index, size);
@@ -811,15 +804,15 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
    // Classic cached path
    const bool sync_cached = SynchronizeBuffer(buffer, cpu_addr, size);
    if (sync_cached) {
-        ++uniform_cache_hits[0];
+        ++channel_state->uniform_cache_hits[0];
    }
-    ++uniform_cache_shots[0];
+    ++channel_state->uniform_cache_shots[0];

    // Skip binding if it's not needed and if the bound buffer is not the fast version
    // This exists to avoid instances where the fast buffer is bound and a GPU write happens
    needs_bind |= HasFastUniformBufferBound(stage, binding_index);
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-        needs_bind |= uniform_buffer_binding_sizes[stage][binding_index] != size;
+        needs_bind |= channel_state->uniform_buffer_binding_sizes[stage][binding_index] != size;
    }
    if (!needs_bind) {
        return;
@@ -827,14 +820,14 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
    const u32 offset = buffer.Offset(cpu_addr);
    if constexpr (IS_OPENGL) {
        // Fast buffer will be unbound
-        fast_bound_uniform_buffers[stage] &= ~(1U << binding_index);
+        channel_state->fast_bound_uniform_buffers[stage] &= ~(1U << binding_index);

        // Mark the index as dirty if offset doesn't match
        const bool is_copy_bind = offset != 0 && !runtime.SupportsNonZeroUniformOffset();
-        dirty_uniform_buffers[stage] |= (is_copy_bind ? 1U : 0U) << index;
+        channel_state->dirty_uniform_buffers[stage] |= (is_copy_bind ? 1U : 0U) << index;
    }
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-        uniform_buffer_binding_sizes[stage][binding_index] = size;
+        channel_state->uniform_buffer_binding_sizes[stage][binding_index] = size;
    }
    if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
        runtime.BindUniformBuffer(stage, binding_index, buffer, offset, size);
@@ -846,15 +839,15 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
 template <class P>
 void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
    u32 binding_index = 0;
-    ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
-        const Binding& binding = storage_buffers[stage][index];
+    ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
+        const Binding& binding = channel_state->storage_buffers[stage][index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        TouchBuffer(buffer, binding.buffer_id);
        const u32 size = binding.size;
        SynchronizeBuffer(buffer, binding.cpu_addr, size);

        const u32 offset = buffer.Offset(binding.cpu_addr);
-        const bool is_written = ((written_storage_buffers[stage] >> index) & 1) != 0;
+        const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
        if constexpr (NEEDS_BIND_STORAGE_INDEX) {
            runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
            ++binding_index;
@@ -866,8 +859,8 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {

 template <class P>
 void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
-    ForEachEnabledBit(enabled_texture_buffers[stage], [&](u32 index) {
-        const TextureBufferBinding& binding = texture_buffers[stage][index];
+    ForEachEnabledBit(channel_state->enabled_texture_buffers[stage], [&](u32 index) {
+        const TextureBufferBinding& binding = channel_state->texture_buffers[stage][index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        const u32 size = binding.size;
        SynchronizeBuffer(buffer, binding.cpu_addr, size);
@@ -875,7 +868,7 @@ void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
        const u32 offset = buffer.Offset(binding.cpu_addr);
        const PixelFormat format = binding.format;
        if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
-            if (((image_texture_buffers[stage] >> index) & 1) != 0) {
+            if (((channel_state->image_texture_buffers[stage] >> index) & 1) != 0) {
                runtime.BindImageBuffer(buffer, offset, size, format);
            } else {
                runtime.BindTextureBuffer(buffer, offset, size, format);
@@ -892,7 +885,7 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
        return;
    }
    for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
-        const Binding& binding = transform_feedback_buffers[index];
+        const Binding& binding = channel_state->transform_feedback_buffers[index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        TouchBuffer(buffer, binding.buffer_id);
        const u32 size = binding.size;
@@ -907,15 +900,16 @@ template <class P>
 void BufferCache<P>::BindHostComputeUniformBuffers() {
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
        // Mark all uniform buffers as dirty
-        dirty_uniform_buffers.fill(~u32{0});
-        fast_bound_uniform_buffers.fill(0);
+        channel_state->dirty_uniform_buffers.fill(~u32{0});
+        channel_state->fast_bound_uniform_buffers.fill(0);
    }
    u32 binding_index = 0;
-    ForEachEnabledBit(enabled_compute_uniform_buffer_mask, [&](u32 index) {
-        const Binding& binding = compute_uniform_buffers[index];
+    ForEachEnabledBit(channel_state->enabled_compute_uniform_buffer_mask, [&](u32 index) {
+        const Binding& binding = channel_state->compute_uniform_buffers[index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        TouchBuffer(buffer, binding.buffer_id);
-        const u32 size = std::min(binding.size, (*compute_uniform_buffer_sizes)[index]);
+        const u32 size =
+            std::min(binding.size, (*channel_state->compute_uniform_buffer_sizes)[index]);
        SynchronizeBuffer(buffer, binding.cpu_addr, size);

        const u32 offset = buffer.Offset(binding.cpu_addr);
@@ -931,15 +925,16 @@ void BufferCache<P>::BindHostComputeUniformBuffers() {
 template <class P>
 void BufferCache<P>::BindHostComputeStorageBuffers() {
    u32 binding_index = 0;
-    ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
-        const Binding& binding = compute_storage_buffers[index];
+    ForEachEnabledBit(channel_state->enabled_compute_storage_buffers, [&](u32 index) {
+        const Binding& binding = channel_state->compute_storage_buffers[index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        TouchBuffer(buffer, binding.buffer_id);
        const u32 size = binding.size;
        SynchronizeBuffer(buffer, binding.cpu_addr, size);

        const u32 offset = buffer.Offset(binding.cpu_addr);
-        const bool is_written = ((written_compute_storage_buffers >> index) & 1) != 0;
+        const bool is_written =
+            ((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
        if constexpr (NEEDS_BIND_STORAGE_INDEX) {
            runtime.BindComputeStorageBuffer(binding_index, buffer, offset, size, is_written);
            ++binding_index;
@@ -951,8 +946,8 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {

 template <class P>
 void BufferCache<P>::BindHostComputeTextureBuffers() {
-    ForEachEnabledBit(enabled_compute_texture_buffers, [&](u32 index) {
-        const TextureBufferBinding& binding = compute_texture_buffers[index];
+    ForEachEnabledBit(channel_state->enabled_compute_texture_buffers, [&](u32 index) {
+        const TextureBufferBinding& binding = channel_state->compute_texture_buffers[index];
        Buffer& buffer = slot_buffers[binding.buffer_id];
        const u32 size = binding.size;
        SynchronizeBuffer(buffer, binding.cpu_addr, size);
@@ -960,7 +955,7 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
        const u32 offset = buffer.Offset(binding.cpu_addr);
        const PixelFormat format = binding.format;
        if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
-            if (((image_compute_texture_buffers >> index) & 1) != 0) {
+            if (((channel_state->image_compute_texture_buffers >> index) & 1) != 0) {
                runtime.BindImageBuffer(buffer, offset, size, format);
            } else {
                runtime.BindTextureBuffer(buffer, offset, size, format);
@@ -974,7 +969,7 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
 template <class P>
 void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
    do {
-        has_deleted_buffers = false;
+        channel_state->has_deleted_buffers = false;
        if (is_indexed) {
            UpdateIndexBuffer();
        }
@@ -988,7 +983,7 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
        if (current_draw_indirect) {
            UpdateDrawIndirect();
        }
-    } while (has_deleted_buffers);
+    } while (channel_state->has_deleted_buffers);
 }

 template <class P>
@@ -1019,7 +1014,7 @@ void BufferCache<P>::UpdateIndexBuffer() {
            slot_buffers.erase(inline_buffer_id);
            inline_buffer_id = CreateBuffer(0, buffer_size);
        }
-        index_buffer = Binding{
+        channel_state->index_buffer = Binding{
            .cpu_addr = 0,
            .size = inline_index_size,
            .buffer_id = inline_buffer_id,
@@ -1035,10 +1030,10 @@ void BufferCache<P>::UpdateIndexBuffer() {
        (index_buffer_ref.count + index_buffer_ref.first) * index_buffer_ref.FormatSizeInBytes();
    const u32 size = std::min(address_size, draw_size);
    if (size == 0 || !cpu_addr) {
-        index_buffer = NULL_BINDING;
+        channel_state->index_buffer = NULL_BINDING;
        return;
    }
-    index_buffer = Binding{
+    channel_state->index_buffer = Binding{
        .cpu_addr = *cpu_addr,
        .size = size,
        .buffer_id = FindBuffer(*cpu_addr, size),
@@ -1071,13 +1066,13 @@ void BufferCache<P>::UpdateVertexBuffer(u32 index) {
    const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
    u32 size = address_size; // TODO: Analyze stride and number of vertices
    if (array.enable == 0 || size == 0 || !cpu_addr) {
-        vertex_buffers[index] = NULL_BINDING;
+        channel_state->vertex_buffers[index] = NULL_BINDING;
        return;
    }
    if (!gpu_memory->IsWithinGPUAddressRange(gpu_addr_end)) {
        size = static_cast<u32>(gpu_memory->MaxContinuousRange(gpu_addr_begin, size));
    }
-    vertex_buffers[index] = Binding{
+    channel_state->vertex_buffers[index] = Binding{
        .cpu_addr = *cpu_addr,
        .size = size,
        .buffer_id = FindBuffer(*cpu_addr, size),
@@ -1099,23 +1094,24 @@ void BufferCache<P>::UpdateDrawIndirect() {
        };
    };
    if (current_draw_indirect->include_count) {
-        update(current_draw_indirect->count_start_address, sizeof(u32), count_buffer_binding);
+        update(current_draw_indirect->count_start_address, sizeof(u32),
+               channel_state->count_buffer_binding);
    }
    update(current_draw_indirect->indirect_start_address, current_draw_indirect->buffer_size,
-           indirect_buffer_binding);
+           channel_state->indirect_buffer_binding);
 }

 template <class P>
 void BufferCache<P>::UpdateUniformBuffers(size_t stage) {
-    ForEachEnabledBit(enabled_uniform_buffer_masks[stage], [&](u32 index) {
-        Binding& binding = uniform_buffers[stage][index];
+    ForEachEnabledBit(channel_state->enabled_uniform_buffer_masks[stage], [&](u32 index) {
+        Binding& binding = channel_state->uniform_buffers[stage][index];
        if (binding.buffer_id) {
            // Already updated
            return;
        }
        // Mark as dirty
        if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-            dirty_uniform_buffers[stage] |= 1U << index;
+            channel_state->dirty_uniform_buffers[stage] |= 1U << index;
        }
        // Resolve buffer
        binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
@@ -1124,10 +1120,10 @@ void BufferCache<P>::UpdateUniformBuffers(size_t stage) {

 template <class P>
 void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
-    const u32 written_mask = written_storage_buffers[stage];
-    ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
+    const u32 written_mask = channel_state->written_storage_buffers[stage];
+    ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
        // Resolve buffer
-        Binding& binding = storage_buffers[stage][index];
+        Binding& binding = channel_state->storage_buffers[stage][index];
        const BufferId buffer_id = FindBuffer(binding.cpu_addr, binding.size);
        binding.buffer_id = buffer_id;
        // Mark buffer as written if needed
@@ -1139,11 +1135,11 @@ void BufferCache<P>::UpdateStorageBuffers(size_t stage) {

 template <class P>
 void BufferCache<P>::UpdateTextureBuffers(size_t stage) {
-    ForEachEnabledBit(enabled_texture_buffers[stage], [&](u32 index) {
-        Binding& binding = texture_buffers[stage][index];
+    ForEachEnabledBit(channel_state->enabled_texture_buffers[stage], [&](u32 index) {
+        Binding& binding = channel_state->texture_buffers[stage][index];
        binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
        // Mark buffer as written if needed
-        if (((written_texture_buffers[stage] >> index) & 1) != 0) {
+        if (((channel_state->written_texture_buffers[stage] >> index) & 1) != 0) {
            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
        }
    });
@@ -1166,11 +1162,11 @@ void BufferCache<P>::UpdateTransformFeedbackBuffer(u32 index) {
    const u32 size = binding.size;
    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
    if (binding.enable == 0 || size == 0 || !cpu_addr) {
-        transform_feedback_buffers[index] = NULL_BINDING;
+        channel_state->transform_feedback_buffers[index] = NULL_BINDING;
        return;
    }
    const BufferId buffer_id = FindBuffer(*cpu_addr, size);
-    transform_feedback_buffers[index] = Binding{
+    channel_state->transform_feedback_buffers[index] = Binding{
        .cpu_addr = *cpu_addr,
        .size = size,
        .buffer_id = buffer_id,
@@ -1180,8 +1176,8 @@ void BufferCache<P>::UpdateTransformFeedbackBuffer(u32 index) {

 template <class P>
 void BufferCache<P>::UpdateComputeUniformBuffers() {
-    ForEachEnabledBit(enabled_compute_uniform_buffer_mask, [&](u32 index) {
-        Binding& binding = compute_uniform_buffers[index];
+    ForEachEnabledBit(channel_state->enabled_compute_uniform_buffer_mask, [&](u32 index) {
+        Binding& binding = channel_state->compute_uniform_buffers[index];
        binding = NULL_BINDING;
        const auto& launch_desc = kepler_compute->launch_description;
        if (((launch_desc.const_buffer_enable_mask >> index) & 1) != 0) {
@@ -1198,12 +1194,12 @@ void BufferCache<P>::UpdateComputeUniformBuffers() {

 template <class P>
 void BufferCache<P>::UpdateComputeStorageBuffers() {
-    ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
+    ForEachEnabledBit(channel_state->enabled_compute_storage_buffers, [&](u32 index) {
        // Resolve buffer
-        Binding& binding = compute_storage_buffers[index];
+        Binding& binding = channel_state->compute_storage_buffers[index];
        binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
        // Mark as written if needed
-        if (((written_compute_storage_buffers >> index) & 1) != 0) {
+        if (((channel_state->written_compute_storage_buffers >> index) & 1) != 0) {
            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
        }
    });
@@ -1211,11 +1207,11 @@ void BufferCache<P>::UpdateComputeStorageBuffers() {

 template <class P>
 void BufferCache<P>::UpdateComputeTextureBuffers() {
-    ForEachEnabledBit(enabled_compute_texture_buffers, [&](u32 index) {
-        Binding& binding = compute_texture_buffers[index];
+    ForEachEnabledBit(channel_state->enabled_compute_texture_buffers, [&](u32 index) {
+        Binding& binding = channel_state->compute_texture_buffers[index];
        binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
        // Mark as written if needed
-        if (((written_compute_texture_buffers >> index) & 1) != 0) {
+        if (((channel_state->written_compute_texture_buffers >> index) & 1) != 0) {
            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
        }
    });
@@ -1223,16 +1219,14 @@ void BufferCache<P>::UpdateComputeTextureBuffers() {

 template <class P>
 void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size) {
-    memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);
-
    if (memory_tracker.IsRegionCpuModified(cpu_addr, size)) {
        SynchronizeBuffer(slot_buffers[buffer_id], cpu_addr, size);
    }
+    memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);

    const IntervalType base_interval{cpu_addr, cpu_addr + size};
    common_ranges.add(base_interval);
    uncommitted_ranges.add(base_interval);
-    pending_ranges.add(base_interval);
 }

 template <class P>
@@ -1284,7 +1278,7 @@ typename BufferCache<P>::OverlapResult BufferCache<P>::ResolveOverlaps(VAddr cpu
        const VAddr overlap_cpu_addr = overlap.CpuAddr();
        const bool expands_left = overlap_cpu_addr < begin;
        if (expands_left) {
-            cpu_addr = begin = overlap_cpu_addr;
+            begin = overlap_cpu_addr;
        }
        const VAddr overlap_end = overlap_cpu_addr + overlap.SizeBytes();
        const bool expands_right = overlap_end > end;
@@ -1298,7 +1292,7 @@ typename BufferCache<P>::OverlapResult BufferCache<P>::ResolveOverlaps(VAddr cpu
            has_stream_leap = true;
            if (expands_right) {
                begin -= CACHING_PAGESIZE * 256;
-                cpu_addr = begin;
+                cpu_addr = begin - CACHING_PAGESIZE;
            }
            if (expands_left) {
                end += CACHING_PAGESIZE * 256;
@@ -1321,7 +1315,7 @@ void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
    if (accumulate_stream_score) {
        new_buffer.IncreaseStreamScore(overlap.StreamScore() + 1);
    }
-    boost::container::small_vector<BufferCopy, 1> copies;
+    boost::container::small_vector<BufferCopy, 10> copies;
    const size_t dst_base_offset = overlap.CpuAddr() - new_buffer.CpuAddr();
    copies.push_back(BufferCopy{
        .src_offset = 0,
@@ -1632,13 +1626,13 @@ void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
    const auto replace = [scalar_replace](std::span<Binding> bindings) {
        std::ranges::for_each(bindings, scalar_replace);
    };
-    scalar_replace(index_buffer);
-    replace(vertex_buffers);
-    std::ranges::for_each(uniform_buffers, replace);
-    std::ranges::for_each(storage_buffers, replace);
-    replace(transform_feedback_buffers);
-    replace(compute_uniform_buffers);
-    replace(compute_storage_buffers);
+    scalar_replace(channel_state->index_buffer);
+    replace(channel_state->vertex_buffers);
+    std::ranges::for_each(channel_state->uniform_buffers, replace);
+    std::ranges::for_each(channel_state->storage_buffers, replace);
+    replace(channel_state->transform_feedback_buffers);
+    replace(channel_state->compute_uniform_buffers);
+    replace(channel_state->compute_storage_buffers);

    // Mark the whole buffer as CPU written to stop tracking CPU writes
    if (!do_not_mark) {
@@ -1656,8 +1650,8 @@ void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
 template <class P>
 void BufferCache<P>::NotifyBufferDeletion() {
    if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
-        dirty_uniform_buffers.fill(~u32{0});
-        uniform_buffer_binding_sizes.fill({});
+        channel_state->dirty_uniform_buffers.fill(~u32{0});
+        channel_state->uniform_buffer_binding_sizes.fill({});
    }
    auto& flags = maxwell3d->dirty.flags;
    flags[Dirty::IndexBuffer] = true;
@@ -1665,27 +1659,27 @@ void BufferCache<P>::NotifyBufferDeletion() {
    for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
        flags[Dirty::VertexBuffer0 + index] = true;
    }
-    has_deleted_buffers = true;
+    channel_state->has_deleted_buffers = true;
 }

 template <class P>
-typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr ssbo_addr,
-                                                                      u32 cbuf_index,
-                                                                      bool is_written) const {
+Binding BufferCache<P>::StorageBufferBinding(GPUVAddr ssbo_addr, u32 cbuf_index,
+                                             bool is_written) const {
    const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
    const auto size = [&]() {
        const bool is_nvn_cbuf = cbuf_index == 0;
        // The NVN driver buffer (index 0) is known to pack the SSBO address followed by its size.
        if (is_nvn_cbuf) {
-            return gpu_memory->Read<u32>(ssbo_addr + 8);
+            const u32 ssbo_size = gpu_memory->Read<u32>(ssbo_addr + 8);
+            if (ssbo_size != 0) {
+                return ssbo_size;
+            }
        }
        // Other titles (notably Doom Eternal) may use STG/LDG on buffer addresses in custom defined
        // cbufs, which do not store the sizes adjacent to the addresses, so use the fully
        // mapped buffer size for now.
        const u32 memory_layout_size = static_cast<u32>(gpu_memory->GetMemoryLayoutSize(gpu_addr));
-        LOG_INFO(HW_GPU, "Binding storage buffer for cbuf index {}, MemoryLayoutSize 0x{:X}",
-                 cbuf_index, memory_layout_size);
-        return memory_layout_size;
+        return std::min(memory_layout_size, static_cast<u32>(8_MiB));
    }();
    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
    if (!cpu_addr || size == 0) {
@@ -1702,8 +1696,8 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
 }

 template <class P>
-typename BufferCache<P>::TextureBufferBinding BufferCache<P>::GetTextureBufferBinding(
-    GPUVAddr gpu_addr, u32 size, PixelFormat format) {
+TextureBufferBinding BufferCache<P>::GetTextureBufferBinding(GPUVAddr gpu_addr, u32 size,
+                                                             PixelFormat format) {
    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
    TextureBufferBinding binding;
    if (!cpu_addr || size == 0) {
@@ -1742,7 +1736,7 @@ std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
 template <class P>
 bool BufferCache<P>::HasFastUniformBufferBound(size_t stage, u32 binding_index) const noexcept {
    if constexpr (IS_OPENGL) {
-        return ((fast_bound_uniform_buffers[stage] >> binding_index) & 1) != 0;
+        return ((channel_state->fast_bound_uniform_buffers[stage] >> binding_index) & 1) != 0;
    } else {
        // Only OpenGL has fast uniform buffers
        return false;
@@ -1751,14 +1745,14 @@ bool BufferCache<P>::HasFastUniformBufferBound(size_t stage, u32 binding_index)

 template <class P>
 std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectCount() {
-    auto& buffer = slot_buffers[count_buffer_binding.buffer_id];
-    return std::make_pair(&buffer, buffer.Offset(count_buffer_binding.cpu_addr));
+    auto& buffer = slot_buffers[channel_state->count_buffer_binding.buffer_id];
+    return std::make_pair(&buffer, buffer.Offset(channel_state->count_buffer_binding.cpu_addr));
 }

 template <class P>
 std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectBuffer() {
-    auto& buffer = slot_buffers[indirect_buffer_binding.buffer_id];
-    return std::make_pair(&buffer, buffer.Offset(indirect_buffer_binding.cpu_addr));
+    auto& buffer = slot_buffers[channel_state->indirect_buffer_binding.buffer_id];
+    return std::make_pair(&buffer, buffer.Offset(channel_state->indirect_buffer_binding.cpu_addr));
 }

 } // namespace VideoCommon
@@ -86,8 +86,78 @@ enum class ObtainBufferOperation : u32 {
    MarkQuery = 3,
 };

-template <typename P>
-class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
+static constexpr BufferId NULL_BUFFER_ID{0};
+static constexpr u32 DEFAULT_SKIP_CACHE_SIZE = static_cast<u32>(4_KiB);
+
+struct Binding {
+    VAddr cpu_addr{};
+    u32 size{};
+    BufferId buffer_id;
+};
+
+struct TextureBufferBinding : Binding {
+    PixelFormat format;
+};
+
+static constexpr Binding NULL_BINDING{
+    .cpu_addr = 0,
+    .size = 0,
+    .buffer_id = NULL_BUFFER_ID,
+};
+
+class BufferCacheChannelInfo : public ChannelInfo {
+public:
+    BufferCacheChannelInfo() = delete;
+    BufferCacheChannelInfo(Tegra::Control::ChannelState& state) noexcept : ChannelInfo(state) {}
+    BufferCacheChannelInfo(const BufferCacheChannelInfo& state) = delete;
+    BufferCacheChannelInfo& operator=(const BufferCacheChannelInfo&) = delete;
+
+    Binding index_buffer;
+    std::array<Binding, NUM_VERTEX_BUFFERS> vertex_buffers;
+    std::array<std::array<Binding, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES> uniform_buffers;
+    std::array<std::array<Binding, NUM_STORAGE_BUFFERS>, NUM_STAGES> storage_buffers;
+    std::array<std::array<TextureBufferBinding, NUM_TEXTURE_BUFFERS>, NUM_STAGES> texture_buffers;
+    std::array<Binding, NUM_TRANSFORM_FEEDBACK_BUFFERS> transform_feedback_buffers;
+    Binding count_buffer_binding;
+    Binding indirect_buffer_binding;
+
+    std::array<Binding, NUM_COMPUTE_UNIFORM_BUFFERS> compute_uniform_buffers;
+    std::array<Binding, NUM_STORAGE_BUFFERS> compute_storage_buffers;
+    std::array<TextureBufferBinding, NUM_TEXTURE_BUFFERS> compute_texture_buffers;
+
+    std::array<u32, NUM_STAGES> enabled_uniform_buffer_masks{};
+    u32 enabled_compute_uniform_buffer_mask = 0;
+
+    const UniformBufferSizes* uniform_buffer_sizes{};
+    const ComputeUniformBufferSizes* compute_uniform_buffer_sizes{};
+
+    std::array<u32, NUM_STAGES> enabled_storage_buffers{};
+    std::array<u32, NUM_STAGES> written_storage_buffers{};
+    u32 enabled_compute_storage_buffers = 0;
+    u32 written_compute_storage_buffers = 0;
+
+    std::array<u32, NUM_STAGES> enabled_texture_buffers{};
+    std::array<u32, NUM_STAGES> written_texture_buffers{};
+    std::array<u32, NUM_STAGES> image_texture_buffers{};
+    u32 enabled_compute_texture_buffers = 0;
+    u32 written_compute_texture_buffers = 0;
+    u32 image_compute_texture_buffers = 0;
+
+    std::array<u32, 16> uniform_cache_hits{};
+    std::array<u32, 16> uniform_cache_shots{};
+
+    u32 uniform_buffer_skip_cache_size = DEFAULT_SKIP_CACHE_SIZE;
+
+    bool has_deleted_buffers = false;
+
+    std::array<u32, NUM_STAGES> dirty_uniform_buffers{};
+    std::array<u32, NUM_STAGES> fast_bound_uniform_buffers{};
+    std::array<std::array<u32, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES>
+        uniform_buffer_binding_sizes{};
+};
+
+template <class P>
+class BufferCache : public VideoCommon::ChannelSetupCaches<BufferCacheChannelInfo> {
    // Page size for caching purposes.
    // This is unrelated to the CPU page size and it can be changed as it seems optimal.
    static constexpr u32 CACHING_PAGEBITS = 16;
@@ -104,8 +174,6 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelI
    static constexpr bool SEPARATE_IMAGE_BUFFERS_BINDINGS = P::SEPARATE_IMAGE_BUFFER_BINDINGS;
    static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = P::IMPLEMENTS_ASYNC_DOWNLOADS;

-    static constexpr BufferId NULL_BUFFER_ID{0};
-
    static constexpr s64 DEFAULT_EXPECTED_MEMORY = 512_MiB;
    static constexpr s64 DEFAULT_CRITICAL_MEMORY = 1_GiB;
    static constexpr s64 TARGET_THRESHOLD = 4_GiB;
@@ -149,8 +217,6 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelI
    using OverlapSection = boost::icl::inter_section<int>;
    using OverlapCounter = boost::icl::split_interval_map<VAddr, int>;

-    struct Empty {};
-
    struct OverlapResult {
        std::vector<BufferId> ids;
        VAddr begin;
@@ -158,25 +224,7 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelI
        bool has_stream_leap = false;
    };

-    struct Binding {
-        VAddr cpu_addr{};
-        u32 size{};
-        BufferId buffer_id;
-    };
-
-    struct TextureBufferBinding : Binding {
-        PixelFormat format;
-    };
-
-    static constexpr Binding NULL_BINDING{
-        .cpu_addr = 0,
-        .size = 0,
-        .buffer_id = NULL_BUFFER_ID,
-    };
-
 public:
-    static constexpr u32 DEFAULT_SKIP_CACHE_SIZE = static_cast<u32>(4_KiB);
-
    explicit BufferCache(VideoCore::RasterizerInterface& rasterizer_,
                         Core::Memory::Memory& cpu_memory_, Runtime& runtime_);

@@ -381,8 +429,6 @@ private:

    void RunGarbageCollector();

-    void WaitOnAsyncFlushes(VAddr cpu_addr, u64 size);
-
    void BindHostIndexBuffer();

    void BindHostVertexBuffers();
@@ -498,56 +544,10 @@ private:

    u32 last_index_count = 0;

-    Binding index_buffer;
-    std::array<Binding, NUM_VERTEX_BUFFERS> vertex_buffers;
-    std::array<std::array<Binding, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES> uniform_buffers;
-    std::array<std::array<Binding, NUM_STORAGE_BUFFERS>, NUM_STAGES> storage_buffers;
-    std::array<std::array<TextureBufferBinding, NUM_TEXTURE_BUFFERS>, NUM_STAGES> texture_buffers;
-    std::array<Binding, NUM_TRANSFORM_FEEDBACK_BUFFERS> transform_feedback_buffers;
-    Binding count_buffer_binding;
-    Binding indirect_buffer_binding;
-
-    std::array<Binding, NUM_COMPUTE_UNIFORM_BUFFERS> compute_uniform_buffers;
-    std::array<Binding, NUM_STORAGE_BUFFERS> compute_storage_buffers;
-    std::array<TextureBufferBinding, NUM_TEXTURE_BUFFERS> compute_texture_buffers;
-
-    std::array<u32, NUM_STAGES> enabled_uniform_buffer_masks{};
-    u32 enabled_compute_uniform_buffer_mask = 0;
-
-    const UniformBufferSizes* uniform_buffer_sizes{};
-    const ComputeUniformBufferSizes* compute_uniform_buffer_sizes{};
-
-    std::array<u32, NUM_STAGES> enabled_storage_buffers{};
-    std::array<u32, NUM_STAGES> written_storage_buffers{};
-    u32 enabled_compute_storage_buffers = 0;
-    u32 written_compute_storage_buffers = 0;
-
-    std::array<u32, NUM_STAGES> enabled_texture_buffers{};
-    std::array<u32, NUM_STAGES> written_texture_buffers{};
-    std::array<u32, NUM_STAGES> image_texture_buffers{};
-    u32 enabled_compute_texture_buffers = 0;
-    u32 written_compute_texture_buffers = 0;
-    u32 image_compute_texture_buffers = 0;
-
-    std::array<u32, 16> uniform_cache_hits{};
-    std::array<u32, 16> uniform_cache_shots{};
-
-    u32 uniform_buffer_skip_cache_size = DEFAULT_SKIP_CACHE_SIZE;
-
-    bool has_deleted_buffers = false;
-
-    std::conditional_t<HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS, std::array<u32, NUM_STAGES>, Empty>
-        dirty_uniform_buffers{};
-    std::conditional_t<IS_OPENGL, std::array<u32, NUM_STAGES>, Empty> fast_bound_uniform_buffers{};
-    std::conditional_t<HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS,
-                       std::array<std::array<u32, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES>, Empty>
-        uniform_buffer_binding_sizes{};
-
    MemoryTracker memory_tracker;
    IntervalSet uncommitted_ranges;
    IntervalSet common_ranges;
    IntervalSet cached_ranges;
-    IntervalSet pending_ranges;
    std::deque<IntervalSet> committed_ranges;

    // Async Buffers
@@ -5,6 +5,7 @@
 #include <fstream>
 #include <vector>
 #include "common/assert.h"
+#include "common/scope_exit.h"
 #include "common/settings.h"
 #include "video_core/host1x/codecs/codec.h"
 #include "video_core/host1x/codecs/h264.h"
@@ -14,6 +15,8 @@
 #include "video_core/memory_manager.h"

 extern "C" {
+#include <libavfilter/buffersink.h>
+#include <libavfilter/buffersrc.h>
 #include <libavutil/opt.h>
 #ifdef LIBVA_FOUND
 // for querying VAAPI driver information
@@ -85,6 +88,10 @@ Codec::~Codec() {
    // Free libav memory
    avcodec_free_context(&av_codec_ctx);
    av_buffer_unref(&av_gpu_decoder);
+
+    if (filters_initialized) {
+        avfilter_graph_free(&av_filter_graph);
+    }
 }

 bool Codec::CreateGpuAvDevice() {
@@ -167,6 +174,62 @@ void Codec::InitializeGpuDecoder() {
    av_codec_ctx->get_format = GetGpuFormat;
 }

+void Codec::InitializeAvFilters(AVFrame* frame) {
+    const AVFilter* buffer_src = avfilter_get_by_name("buffer");
+    const AVFilter* buffer_sink = avfilter_get_by_name("buffersink");
+    AVFilterInOut* inputs = avfilter_inout_alloc();
+    AVFilterInOut* outputs = avfilter_inout_alloc();
+    SCOPE_EXIT({
+        avfilter_inout_free(&inputs);
+        avfilter_inout_free(&outputs);
+    });
+
+    // Don't know how to get the accurate time_base but it doesn't matter for yadif filter
+    // so just use 1/1 to make buffer filter happy
+    std::string args = fmt::format("video_size={}x{}:pix_fmt={}:time_base=1/1", frame->width,
+                                   frame->height, frame->format);
+
+    av_filter_graph = avfilter_graph_alloc();
+    int ret = avfilter_graph_create_filter(&av_filter_src_ctx, buffer_src, "in", args.c_str(),
+                                           nullptr, av_filter_graph);
+    if (ret < 0) {
+        LOG_ERROR(Service_NVDRV, "avfilter_graph_create_filter source error: {}", ret);
+        return;
+    }
+
+    ret = avfilter_graph_create_filter(&av_filter_sink_ctx, buffer_sink, "out", nullptr, nullptr,
+                                       av_filter_graph);
+    if (ret < 0) {
+        LOG_ERROR(Service_NVDRV, "avfilter_graph_create_filter sink error: {}", ret);
+        return;
+    }
+
+    inputs->name = av_strdup("out");
+    inputs->filter_ctx = av_filter_sink_ctx;
+    inputs->pad_idx = 0;
+    inputs->next = nullptr;
+
+    outputs->name = av_strdup("in");
+    outputs->filter_ctx = av_filter_src_ctx;
+    outputs->pad_idx = 0;
+    outputs->next = nullptr;
+
+    const char* description = "yadif=1:-1:0";
+    ret = avfilter_graph_parse_ptr(av_filter_graph, description, &inputs, &outputs, nullptr);
+    if (ret < 0) {
+        LOG_ERROR(Service_NVDRV, "avfilter_graph_parse_ptr error: {}", ret);
+        return;
+    }
+
+    ret = avfilter_graph_config(av_filter_graph, nullptr);
+    if (ret < 0) {
+        LOG_ERROR(Service_NVDRV, "avfilter_graph_config error: {}", ret);
+        return;
+    }
+
+    filters_initialized = true;
+}
+
 void Codec::Initialize() {
    const AVCodecID codec = [&] {
        switch (current_codec) {
@@ -271,8 +334,34 @@ void Codec::Decode() {
        UNIMPLEMENTED_MSG("Unexpected video format: {}", final_frame->format);
        return;
    }
-    av_frames.push(std::move(final_frame));
-    if (av_frames.size() > 10) {
+    if (!final_frame->interlaced_frame) {
+        av_frames.push(std::move(final_frame));
+    } else {
+        if (!filters_initialized) {
+            InitializeAvFilters(final_frame.get());
+        }
+        if (const int ret = av_buffersrc_add_frame_flags(av_filter_src_ctx, final_frame.get(),
+                                                         AV_BUFFERSRC_FLAG_KEEP_REF);
+            ret) {
+            LOG_DEBUG(Service_NVDRV, "av_buffersrc_add_frame_flags error {}", ret);
+            return;
+        }
+        while (true) {
+            auto filter_frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
+
+            int ret = av_buffersink_get_frame(av_filter_sink_ctx, filter_frame.get());
+
+            if (ret == AVERROR(EAGAIN) || ret == AVERROR(AVERROR_EOF))
+                break;
+            if (ret < 0) {
+                LOG_DEBUG(Service_NVDRV, "av_buffersink_get_frame error {}", ret);
+                return;
+            }
+
+            av_frames.push(std::move(filter_frame));
+        }
+    }
+    while (av_frames.size() > 10) {
        LOG_TRACE(Service_NVDRV, "av_frames.push overflow dropped frame");
        av_frames.pop();
    }
@@ -15,6 +15,7 @@ extern "C" {
 #pragma GCC diagnostic ignored "-Wconversion"
 #endif
 #include <libavcodec/avcodec.h>
+#include <libavfilter/avfilter.h>
 #if defined(__GNUC__) || defined(__clang__)
 #pragma GCC diagnostic pop
 #endif
@@ -61,17 +62,24 @@ public:
 private:
    void InitializeAvCodecContext();

+    void InitializeAvFilters(AVFrame* frame);
+
    void InitializeGpuDecoder();

    bool CreateGpuAvDevice();

    bool initialized{};
+    bool filters_initialized{};
    Host1x::NvdecCommon::VideoCodec current_codec{Host1x::NvdecCommon::VideoCodec::None};

    const AVCodec* av_codec{nullptr};
    AVCodecContext* av_codec_ctx{nullptr};
    AVBufferRef* av_gpu_decoder{nullptr};

+    AVFilterContext* av_filter_src_ctx{nullptr};
+    AVFilterContext* av_filter_sink_ctx{nullptr};
+    AVFilterGraph* av_filter_graph{nullptr};
+
    Host1x::Host1x& host1x;
    const Host1x::NvdecCommon::NvdecRegisters& state;
    std::unique_ptr<Decoder::H264> h264_decoder;
--- a/Show More
+++ b/Show More