28 files changed, 661 insertions, 739 deletions
diff --git a/src/common/assert.h b/src/common/assert.h
index b3ba35c0f9..33060d865c 100644
--- a/src/common/assert.h
+++ b/src/common/assert.h
@@ -52,8 +52,12 @@ assert_noinline_call(const Fn& fn) {
 #define DEBUG_ASSERT(_a_) ASSERT(_a_)
 #define DEBUG_ASSERT_MSG(_a_, ...) ASSERT_MSG(_a_, __VA_ARGS__)
 #else // not debug
-#define DEBUG_ASSERT(_a_)
-#define DEBUG_ASSERT_MSG(_a_, _desc_, ...)
+#define DEBUG_ASSERT(_a_)                                                                          \
+    do {                                                                                           \
+    } while (0)
+#define DEBUG_ASSERT_MSG(_a_, _desc_, ...)                                                         \
+    do {                                                                                           \
+    } while (0)
 #endif
 
 #define UNIMPLEMENTED() ASSERT_MSG(false, "Unimplemented code!")
diff --git a/src/common/hex_util.h b/src/common/hex_util.h
index f5f9e45077..5e9b6ef8b5 100644
--- a/src/common/hex_util.h
+++ b/src/common/hex_util.h
@@ -61,7 +61,7 @@ template <typename ContiguousContainer>
     return out;
 }
 
-[[nodiscard]] constexpr std::array<u8, 16> AsArray(const char (&data)[17]) {
+[[nodiscard]] constexpr std::array<u8, 16> AsArray(const char (&data)[33]) {
     return HexStringToArray<16>(data);
 }
 
diff --git a/src/common/uuid.cpp b/src/common/uuid.cpp
index 18303a1e32..d7435a6e95 100644
--- a/src/common/uuid.cpp
+++ b/src/common/uuid.cpp
@@ -6,10 +6,64 @@
 
 #include <fmt/format.h>
 
+#include "common/assert.h"
 #include "common/uuid.h"
 
 namespace Common {
 
+namespace {
+
+bool IsHexDigit(char c) {
+    return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F');
+}
+
+u8 HexCharToByte(char c) {
+    if (c >= '0' && c <= '9') {
+        return static_cast<u8>(c - '0');
+    }
+    if (c >= 'a' && c <= 'f') {
+        return static_cast<u8>(c - 'a' + 10);
+    }
+    if (c >= 'A' && c <= 'F') {
+        return static_cast<u8>(c - 'A' + 10);
+    }
+    ASSERT_MSG(false, "{} is not a hexadecimal digit!", c);
+    return u8{0};
+}
+
+} // Anonymous namespace
+
+u128 HexStringToU128(std::string_view hex_string) {
+    const size_t length = hex_string.length();
+
+    // Detect "0x" prefix.
+    const bool has_0x_prefix = length > 2 && hex_string[0] == '0' && hex_string[1] == 'x';
+    const size_t offset = has_0x_prefix ? 2 : 0;
+
+    // Check length.
+    if (length > 32 + offset) {
+        ASSERT_MSG(false, "hex_string has more than 32 hexadecimal characters!");
+        return INVALID_UUID;
+    }
+
+    u64 lo = 0;
+    u64 hi = 0;
+    for (size_t i = 0; i < length - offset; ++i) {
+        const char c = hex_string[length - 1 - i];
+        if (!IsHexDigit(c)) {
+            ASSERT_MSG(false, "{} is not a hexadecimal digit!", c);
+            return INVALID_UUID;
+        }
+        if (i < 16) {
+            lo |= u64{HexCharToByte(c)} << (i * 4);
+        }
+        if (i >= 16) {
+            hi |= u64{HexCharToByte(c)} << ((i - 16) * 4);
+        }
+    }
+    return u128{lo, hi};
+}
+
 UUID UUID::Generate() {
     std::random_device device;
     std::mt19937 gen(device());
diff --git a/src/common/uuid.h b/src/common/uuid.h
index 0ffa37e7cb..aeb36939a3 100644
--- a/src/common/uuid.h
+++ b/src/common/uuid.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <string>
+#include <string_view>
 
 #include "common/common_types.h"
 
@@ -12,12 +13,30 @@ namespace Common {
 
 constexpr u128 INVALID_UUID{{0, 0}};
 
+/**
+ * Converts a hex string to a 128-bit unsigned integer.
+ *
+ * The hex string can be formatted in lowercase or uppercase, with or without the "0x" prefix.
+ *
+ * This function will assert and return INVALID_UUID under the following conditions:
+ * - If the hex string is more than 32 characters long
+ * - If the hex string contains non-hexadecimal characters
+ *
+ * @param hex_string Hexadecimal string
+ *
+ * @returns A 128-bit unsigned integer if successfully converted, INVALID_UUID otherwise.
+ */
+[[nodiscard]] u128 HexStringToU128(std::string_view hex_string);
+
 struct UUID {
     // UUIDs which are 0 are considered invalid!
     u128 uuid;
     UUID() = default;
     constexpr explicit UUID(const u128& id) : uuid{id} {}
     constexpr explicit UUID(const u64 lo, const u64 hi) : uuid{{lo, hi}} {}
+    explicit UUID(std::string_view hex_string) {
+        uuid = HexStringToU128(hex_string);
+    }
 
     [[nodiscard]] constexpr explicit operator bool() const {
         return uuid != INVALID_UUID;
diff --git a/src/core/hle/service/am/applets/applet_software_keyboard.cpp b/src/core/hle/service/am/applets/applet_software_keyboard.cpp
index 673abb7552..c89aa1bbfc 100644
--- a/src/core/hle/service/am/applets/applet_software_keyboard.cpp
+++ b/src/core/hle/service/am/applets/applet_software_keyboard.cpp
@@ -377,7 +377,8 @@ void SoftwareKeyboard::SubmitForTextCheck(std::u16string submitted_text) {
 
     if (swkbd_config_common.use_utf8) {
         std::string utf8_submitted_text = Common::UTF16ToUTF8(current_text);
-        const u64 buffer_size = utf8_submitted_text.size();
+        // Include the null terminator in the buffer size.
+        const u64 buffer_size = utf8_submitted_text.size() + 1;
 
         LOG_DEBUG(Service_AM, "\nBuffer Size: {}\nUTF-8 Submitted Text: {}", buffer_size,
                   utf8_submitted_text);
@@ -386,7 +387,8 @@ void SoftwareKeyboard::SubmitForTextCheck(std::u16string submitted_text) {
         std::memcpy(out_data.data() + sizeof(u64), utf8_submitted_text.data(),
                     utf8_submitted_text.size());
     } else {
-        const u64 buffer_size = current_text.size() * sizeof(char16_t);
+        // Include the null terminator in the buffer size.
+        const u64 buffer_size = (current_text.size() + 1) * sizeof(char16_t);
 
         LOG_DEBUG(Service_AM, "\nBuffer Size: {}\nUTF-16 Submitted Text: {}", buffer_size,
                   Common::UTF16ToUTF8(current_text));
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
index 1403a39d0f..845de724d1 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
@@ -166,8 +166,6 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
             LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size);
         } else {
             cmd_buffer.map_address = object->dma_map_addr;
-            AddBufferMap(object->dma_map_addr, object->size, object->addr,
-                         object->status == nvmap::Object::Status::Allocated);
         }
     }
     std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
@@ -178,30 +176,11 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
 }
 
 NvResult nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
-    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
-    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
-
-    auto& gpu = system.GPU();
-
-    for (auto& cmd_buffer : cmd_buffer_handles) {
-        const auto object{nvmap_dev->GetObject(cmd_buffer.map_handle)};
-        if (!object) {
-            LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmd_buffer.map_handle);
-            std::memcpy(output.data(), &params, output.size());
-            return NvResult::InvalidState;
-        }
-        if (const auto size{RemoveBufferMap(object->dma_map_addr)}; size) {
-            gpu.MemoryManager().Unmap(object->dma_map_addr, *size);
-        } else {
-            // This occurs quite frequently, however does not seem to impact functionality
-            LOG_DEBUG(Service_NVDRV, "invalid offset=0x{:X} dma=0x{:X}", object->addr,
-                      object->dma_map_addr);
-        }
-        object->dma_map_addr = 0;
-    }
+    // This is intntionally stubbed.
+    // Skip unmapping buffers here, as to not break the continuity of the VP9 reference frame
+    // addresses, and risk invalidating data before the async GPU thread is done with it
     std::memset(output.data(), 0, output.size());
+    LOG_DEBUG(Service_NVDRV, "(STUBBED) called");
     return NvResult::Success;
 }
 
@@ -212,33 +191,4 @@ NvResult nvhost_nvdec_common::SetSubmitTimeout(const std::vector<u8>& input,
     return NvResult::Success;
 }
 
-std::optional<nvhost_nvdec_common::BufferMap> nvhost_nvdec_common::FindBufferMap(
-    GPUVAddr gpu_addr) const {
-    const auto it = std::find_if(
-        buffer_mappings.begin(), buffer_mappings.upper_bound(gpu_addr), [&](const auto& entry) {
-            return (gpu_addr >= entry.second.StartAddr() && gpu_addr < entry.second.EndAddr());
-        });
-
-    ASSERT(it != buffer_mappings.end());
-    return it->second;
-}
-
-void nvhost_nvdec_common::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr,
-                                       bool is_allocated) {
-    buffer_mappings.insert_or_assign(gpu_addr, BufferMap{gpu_addr, size, cpu_addr, is_allocated});
-}
-
-std::optional<std::size_t> nvhost_nvdec_common::RemoveBufferMap(GPUVAddr gpu_addr) {
-    const auto iter{buffer_mappings.find(gpu_addr)};
-    if (iter == buffer_mappings.end()) {
-        return std::nullopt;
-    }
-    std::size_t size = 0;
-    if (iter->second.IsAllocated()) {
-        size = iter->second.Size();
-    }
-    buffer_mappings.erase(iter);
-    return size;
-}
-
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
index da10f5f412..af59f00d2a 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h
@@ -23,45 +23,6 @@ public:
     ~nvhost_nvdec_common() override;
 
 protected:
-    class BufferMap final {
-    public:
-        constexpr BufferMap() = default;
-
-        constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_)
-            : start_addr{start_addr_}, end_addr{start_addr_ + size_} {}
-
-        constexpr BufferMap(GPUVAddr start_addr_, std::size_t size_, VAddr cpu_addr_,
-                            bool is_allocated_)
-            : start_addr{start_addr_}, end_addr{start_addr_ + size_}, cpu_addr{cpu_addr_},
-              is_allocated{is_allocated_} {}
-
-        constexpr VAddr StartAddr() const {
-            return start_addr;
-        }
-
-        constexpr VAddr EndAddr() const {
-            return end_addr;
-        }
-
-        constexpr std::size_t Size() const {
-            return end_addr - start_addr;
-        }
-
-        constexpr VAddr CpuAddr() const {
-            return cpu_addr;
-        }
-
-        constexpr bool IsAllocated() const {
-            return is_allocated;
-        }
-
-    private:
-        GPUVAddr start_addr{};
-        GPUVAddr end_addr{};
-        VAddr cpu_addr{};
-        bool is_allocated{};
-    };
-
     struct IoctlSetNvmapFD {
         s32_le nvmap_fd{};
     };
@@ -154,17 +115,11 @@ protected:
     NvResult UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output);
     NvResult SetSubmitTimeout(const std::vector<u8>& input, std::vector<u8>& output);
 
-    std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const;
-    void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated);
-    std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr);
-
     s32_le nvmap_fd{};
     u32_le submit_timeout{};
     std::shared_ptr<nvmap> nvmap_dev;
     SyncpointManager& syncpoint_manager;
     std::array<u32, MaxSyncPoints> device_syncpoints{};
-    // This is expected to be ordered, therefore we must use a map, not unordered_map
-    std::map<GPUVAddr, BufferMap> buffer_mappings;
 };
 }; // namespace Devices
 } // namespace Service::Nvidia
diff --git a/src/core/network/network.cpp b/src/core/network/network.cpp
index 526bfa1108..375bc79ec7 100644
--- a/src/core/network/network.cpp
+++ b/src/core/network/network.cpp
@@ -570,7 +570,7 @@ std::pair<s32, Errno> Socket::SendTo(u32 flags, const std::vector<u8>& message,
     ASSERT(flags == 0);
 
     const sockaddr* to = nullptr;
-    const int tolen = addr ? 0 : sizeof(sockaddr);
+    const int tolen = addr ? sizeof(sockaddr) : 0;
     sockaddr host_addr_in;
 
     if (addr) {
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 333f6f35fd..1eb67c051b 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,5 +1,10 @@
 add_subdirectory(host_shaders)
 
+if(LIBVA_FOUND)
+    set_source_files_properties(command_classes/codecs/codec.cpp
+        PROPERTIES COMPILE_DEFINITIONS LIBVA_FOUND=1)
+endif()
+
 add_library(video_core STATIC
     buffer_cache/buffer_base.h
     buffer_cache/buffer_cache.cpp
diff --git a/src/video_core/command_classes/codecs/codec.cpp b/src/video_core/command_classes/codecs/codec.cpp
index 1b4bbc8ac4..f798a00531 100644
--- a/src/video_core/command_classes/codecs/codec.cpp
+++ b/src/video_core/command_classes/codecs/codec.cpp
@@ -2,7 +2,6 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <cstring>
 #include <fstream>
 #include <vector>
 #include "common/assert.h"
@@ -17,10 +16,47 @@ extern "C" {
 }
 
 namespace Tegra {
+#if defined(LIBVA_FOUND)
+// Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c originally under MIT license
+namespace {
+constexpr std::array<const char*, 2> VAAPI_DRIVERS = {
+    "i915",
+    "amdgpu",
+};
+
+AVPixelFormat GetHwFormat(AVCodecContext*, const AVPixelFormat* pix_fmts) {
+    for (const AVPixelFormat* p = pix_fmts; *p != AV_PIX_FMT_NONE; ++p) {
+        if (*p == AV_PIX_FMT_VAAPI) {
+            return AV_PIX_FMT_VAAPI;
+        }
+    }
+    LOG_INFO(Service_NVDRV, "Could not find compatible GPU AV format, falling back to CPU");
+    return *pix_fmts;
+}
+
+bool CreateVaapiHwdevice(AVBufferRef** av_hw_device) {
+    AVDictionary* hwdevice_options = nullptr;
+    av_dict_set(&hwdevice_options, "connection_type", "drm", 0);
+    for (const auto& driver : VAAPI_DRIVERS) {
+        av_dict_set(&hwdevice_options, "kernel_driver", driver, 0);
+        const int hwdevice_error = av_hwdevice_ctx_create(av_hw_device, AV_HWDEVICE_TYPE_VAAPI,
+                                                          nullptr, hwdevice_options, 0);
+        if (hwdevice_error >= 0) {
+            LOG_INFO(Service_NVDRV, "Using VA-API with {}", driver);
+            av_dict_free(&hwdevice_options);
+            return true;
+        }
+        LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed {}", hwdevice_error);
+    }
+    LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed for all drivers");
+    av_dict_free(&hwdevice_options);
+    return false;
+}
+} // namespace
+#endif
 
 void AVFrameDeleter(AVFrame* ptr) {
-    av_frame_unref(ptr);
-    av_free(ptr);
+    av_frame_free(&ptr);
 }
 
 Codec::Codec(GPU& gpu_, const NvdecCommon::NvdecRegisters& regs)
@@ -32,19 +68,31 @@ Codec::~Codec() {
         return;
     }
     // Free libav memory
-    AVFrame* av_frame{nullptr};
     avcodec_send_packet(av_codec_ctx, nullptr);
-    av_frame = av_frame_alloc();
+    AVFrame* av_frame = av_frame_alloc();
     avcodec_receive_frame(av_codec_ctx, av_frame);
     avcodec_flush_buffers(av_codec_ctx);
-
-    av_frame_unref(av_frame);
-    av_free(av_frame);
+    av_frame_free(&av_frame);
     avcodec_close(av_codec_ctx);
+    av_buffer_unref(&av_hw_device);
+}
+
+void Codec::InitializeHwdec() {
+    // Prioritize integrated GPU to mitigate bandwidth bottlenecks
+#if defined(LIBVA_FOUND)
+    if (CreateVaapiHwdevice(&av_hw_device)) {
+        const auto hw_device_ctx = av_buffer_ref(av_hw_device);
+        ASSERT_MSG(hw_device_ctx, "av_buffer_ref failed");
+        av_codec_ctx->hw_device_ctx = hw_device_ctx;
+        av_codec_ctx->get_format = GetHwFormat;
+        return;
+    }
+#endif
+    // TODO more GPU accelerated decoders
 }
 
 void Codec::Initialize() {
-    AVCodecID codec{AV_CODEC_ID_NONE};
+    AVCodecID codec;
     switch (current_codec) {
     case NvdecCommon::VideoCodec::H264:
         codec = AV_CODEC_ID_H264;
@@ -53,22 +101,24 @@ void Codec::Initialize() {
         codec = AV_CODEC_ID_VP9;
         break;
     default:
+        UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
         return;
     }
     av_codec = avcodec_find_decoder(codec);
     av_codec_ctx = avcodec_alloc_context3(av_codec);
     av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
-
-    // TODO(ameerj): libavcodec gpu hw acceleration
-
+    InitializeHwdec();
+    if (!av_codec_ctx->hw_device_ctx) {
+        LOG_INFO(Service_NVDRV, "Using FFmpeg software decoding");
+    }
     const auto av_error = avcodec_open2(av_codec_ctx, av_codec, nullptr);
     if (av_error < 0) {
         LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed.");
         avcodec_close(av_codec_ctx);
+        av_buffer_unref(&av_hw_device);
         return;
     }
     initialized = true;
-    return;
 }
 
 void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
@@ -80,36 +130,64 @@ void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) {
 
 void Codec::Decode() {
     const bool is_first_frame = !initialized;
-    if (!initialized) {
+    if (is_first_frame) {
         Initialize();
     }
-
     bool vp9_hidden_frame = false;
-    AVPacket packet{};
-    av_init_packet(&packet);
     std::vector<u8> frame_data;
-
     if (current_codec == NvdecCommon::VideoCodec::H264) {
         frame_data = h264_decoder->ComposeFrameHeader(state, is_first_frame);
     } else if (current_codec == NvdecCommon::VideoCodec::Vp9) {
         frame_data = vp9_decoder->ComposeFrameHeader(state);
         vp9_hidden_frame = vp9_decoder->WasFrameHidden();
     }
-
+    AVPacket packet{};
+    av_init_packet(&packet);
     packet.data = frame_data.data();
     packet.size = static_cast<s32>(frame_data.size());
-
-    avcodec_send_packet(av_codec_ctx, &packet);
-
-    if (!vp9_hidden_frame) {
-        // Only receive/store visible frames
-        AVFramePtr frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
-        avcodec_receive_frame(av_codec_ctx, frame.get());
-        av_frames.push(std::move(frame));
-        // Limit queue to 10 frames. Workaround for ZLA decode and queue spam
-        if (av_frames.size() > 10) {
-            av_frames.pop();
-        }
+    if (const int ret = avcodec_send_packet(av_codec_ctx, &packet); ret) {
+        LOG_DEBUG(Service_NVDRV, "avcodec_send_packet error {}", ret);
+        return;
+    }
+    // Only receive/store visible frames
+    if (vp9_hidden_frame) {
+        return;
+    }
+    AVFrame* hw_frame = av_frame_alloc();
+    AVFrame* sw_frame = hw_frame;
+    ASSERT_MSG(hw_frame, "av_frame_alloc hw_frame failed");
+    if (const int ret = avcodec_receive_frame(av_codec_ctx, hw_frame); ret) {
+        LOG_DEBUG(Service_NVDRV, "avcodec_receive_frame error {}", ret);
+        av_frame_free(&hw_frame);
+        return;
+    }
+    if (!hw_frame->width || !hw_frame->height) {
+        LOG_WARNING(Service_NVDRV, "Zero width or height in frame");
+        av_frame_free(&hw_frame);
+        return;
+    }
+#if defined(LIBVA_FOUND)
+    // Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c under MIT license
+    if (hw_frame->format == AV_PIX_FMT_VAAPI) {
+        sw_frame = av_frame_alloc();
+        ASSERT_MSG(sw_frame, "av_frame_alloc sw_frame failed");
+        // Can't use AV_PIX_FMT_YUV420P and share code with software decoding in vic.cpp
+        // because Intel drivers crash unless using AV_PIX_FMT_NV12
+        sw_frame->format = AV_PIX_FMT_NV12;
+        const int transfer_data_ret = av_hwframe_transfer_data(sw_frame, hw_frame, 0);
+        ASSERT_MSG(!transfer_data_ret, "av_hwframe_transfer_data error {}", transfer_data_ret);
+        av_frame_free(&hw_frame);
+    }
+#endif
+    if (sw_frame->format != AV_PIX_FMT_YUV420P && sw_frame->format != AV_PIX_FMT_NV12) {
+        UNIMPLEMENTED_MSG("Unexpected video format from host graphics: {}", sw_frame->format);
+        av_frame_free(&sw_frame);
+        return;
+    }
+    av_frames.push(AVFramePtr{sw_frame, AVFrameDeleter});
+    if (av_frames.size() > 10) {
+        LOG_TRACE(Service_NVDRV, "av_frames.push overflow dropped frame");
+        av_frames.pop();
     }
 }
 
@@ -119,7 +197,6 @@ AVFramePtr Codec::GetCurrentFrame() {
     if (av_frames.empty()) {
         return AVFramePtr{nullptr, AVFrameDeleter};
     }
-
     AVFramePtr frame = std::move(av_frames.front());
     av_frames.pop();
     return frame;
@@ -144,6 +221,5 @@ std::string_view Codec::GetCurrentCodecName() const {
     default:
         return "Unknown";
     }
-};
-
+}
 } // namespace Tegra
diff --git a/src/video_core/command_classes/codecs/codec.h b/src/video_core/command_classes/codecs/codec.h
index 96c823c765..71936203fb 100644
--- a/src/video_core/command_classes/codecs/codec.h
+++ b/src/video_core/command_classes/codecs/codec.h
@@ -22,7 +22,6 @@ extern "C" {
 
 namespace Tegra {
 class GPU;
-struct VicRegisters;
 
 void AVFrameDeleter(AVFrame* ptr);
 using AVFramePtr = std::unique_ptr<AVFrame, decltype(&AVFrameDeleter)>;
@@ -55,10 +54,13 @@ public:
     [[nodiscard]] std::string_view GetCurrentCodecName() const;
 
 private:
+    void InitializeHwdec();
+
     bool initialized{};
     NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
 
     AVCodec* av_codec{nullptr};
+    AVBufferRef* av_hw_device{nullptr};
     AVCodecContext* av_codec_ctx{nullptr};
 
     GPU& gpu;
diff --git a/src/video_core/command_classes/codecs/vp9.cpp b/src/video_core/command_classes/codecs/vp9.cpp
index 902bc2a982..7eecb39915 100644
--- a/src/video_core/command_classes/codecs/vp9.cpp
+++ b/src/video_core/command_classes/codecs/vp9.cpp
@@ -11,6 +11,9 @@
 
 namespace Tegra::Decoder {
 namespace {
+constexpr u32 diff_update_probability = 252;
+constexpr u32 frame_sync_code = 0x498342;
+
 // Default compressed header probabilities once frame context resets
 constexpr Vp9EntropyProbs default_probs{
     .y_mode_prob{
@@ -361,8 +364,7 @@ Vp9PictureInfo VP9::GetVp9PictureInfo(const NvdecCommon::NvdecRegisters& state)
     InsertEntropy(state.vp9_entropy_probs_offset, vp9_info.entropy);
 
     // surface_luma_offset[0:3] contains the address of the reference frame offsets in the following
-    // order: last, golden, altref, current. It may be worthwhile to track the updates done here
-    // to avoid buffering frame data needed for reference frame updating in the header composition.
+    // order: last, golden, altref, current.
     std::copy(state.surface_luma_offset.begin(), state.surface_luma_offset.begin() + 4,
               vp9_info.frame_offsets.begin());
 
@@ -384,33 +386,18 @@ Vp9FrameContainer VP9::GetCurrentFrame(const NvdecCommon::NvdecRegisters& state)
         gpu.MemoryManager().ReadBlock(state.frame_bitstream_offset, current_frame.bit_stream.data(),
                                       current_frame.info.bitstream_size);
     }
-    // Buffer two frames, saving the last show frame info
-    if (!next_next_frame.bit_stream.empty()) {
+    if (!next_frame.bit_stream.empty()) {
         Vp9FrameContainer temp{
             .info = current_frame.info,
             .bit_stream = std::move(current_frame.bit_stream),
         };
-        next_next_frame.info.show_frame = current_frame.info.last_frame_shown;
-        current_frame.info = next_next_frame.info;
-        current_frame.bit_stream = std::move(next_next_frame.bit_stream);
-        next_next_frame = std::move(temp);
-
-        if (!next_frame.bit_stream.empty()) {
-            Vp9FrameContainer temp2{
-                .info = current_frame.info,
-                .bit_stream = std::move(current_frame.bit_stream),
-            };
-            next_frame.info.show_frame = current_frame.info.last_frame_shown;
-            current_frame.info = next_frame.info;
-            current_frame.bit_stream = std::move(next_frame.bit_stream);
-            next_frame = std::move(temp2);
-        } else {
-            next_frame.info = current_frame.info;
-            next_frame.bit_stream = std::move(current_frame.bit_stream);
-        }
+        next_frame.info.show_frame = current_frame.info.last_frame_shown;
+        current_frame.info = next_frame.info;
+        current_frame.bit_stream = std::move(next_frame.bit_stream);
+        next_frame = std::move(temp);
     } else {
-        next_next_frame.info = current_frame.info;
-        next_next_frame.bit_stream = std::move(current_frame.bit_stream);
+        next_frame.info = current_frame.info;
+        next_frame.bit_stream = std::move(current_frame.bit_stream);
     }
     return current_frame;
 }
@@ -613,86 +600,64 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
 
         // Reset context
         prev_frame_probs = default_probs;
-        swap_next_golden = false;
+        swap_ref_indices = false;
         loop_filter_ref_deltas.fill(0);
         loop_filter_mode_deltas.fill(0);
-
-        // allow frames offsets to stabilize before checking for golden frames
-        grace_period = 4;
-
-        // On key frames, all frame slots are set to the current frame,
-        // so the value of the selected slot doesn't really matter.
-        frame_ctxs.fill({current_frame_number, false, default_probs});
+        frame_ctxs.fill(default_probs);
 
         // intra only, meaning the frame can be recreated with no other references
         current_frame_info.intra_only = true;
-
     } else {
-
         if (!current_frame_info.show_frame) {
             uncomp_writer.WriteBit(current_frame_info.intra_only);
-            if (!current_frame_info.last_frame_was_key) {
-                swap_next_golden = !swap_next_golden;
-            }
         } else {
             current_frame_info.intra_only = false;
         }
         if (!current_frame_info.error_resilient_mode) {
             uncomp_writer.WriteU(0, 2); // Reset frame context.
         }
-
-        // Last, Golden, Altref frames
-        std::array<s32, 3> ref_frame_index{0, 1, 2};
-
-        // Set when next frame is hidden
-        // altref and golden references are swapped
-        if (swap_next_golden) {
-            ref_frame_index = std::array<s32, 3>{0, 2, 1};
+        const auto& curr_offsets = current_frame_info.frame_offsets;
+        const auto& next_offsets = next_frame.info.frame_offsets;
+        const bool ref_frames_different = curr_offsets[1] != curr_offsets[2];
+        const bool next_references_swap =
+            (next_offsets[1] == curr_offsets[2]) || (next_offsets[2] == curr_offsets[1]);
+        const bool needs_ref_swap = ref_frames_different && next_references_swap;
+        if (needs_ref_swap) {
+            swap_ref_indices = !swap_ref_indices;
         }
-
-        // update Last Frame
-        u64 refresh_frame_flags = 1;
-
-        // golden frame may refresh, determined if the next golden frame offset is changed
-        bool golden_refresh = false;
-        if (grace_period <= 0) {
-            for (s32 index = 1; index < 3; ++index) {
-                if (current_frame_info.frame_offsets[index] !=
-                    next_frame.info.frame_offsets[index]) {
-                    current_frame_info.refresh_frame[index] = true;
-                    golden_refresh = true;
-                    grace_period = 3;
-                }
+        union {
+            u32 raw;
+            BitField<0, 1, u32> refresh_last;
+            BitField<1, 2, u32> refresh_golden;
+            BitField<2, 1, u32> refresh_alt;
+        } refresh_frame_flags;
+
+        refresh_frame_flags.raw = 0;
+        for (u32 index = 0; index < 3; ++index) {
+            // Refresh indices that use the current frame as an index
+            if (curr_offsets[3] == next_offsets[index]) {
+                refresh_frame_flags.raw |= 1u << index;
             }
         }
-
-        if (current_frame_info.show_frame &&
-            (!next_frame.info.show_frame || next_frame.info.is_key_frame)) {
-            // Update golden frame
-            refresh_frame_flags = swap_next_golden ? 2 : 4;
-        }
-
-        if (!current_frame_info.show_frame) {
-            // Update altref
-            refresh_frame_flags = swap_next_golden ? 2 : 4;
-        } else if (golden_refresh) {
-            refresh_frame_flags = 3;
+        if (swap_ref_indices) {
+            const u32 temp = refresh_frame_flags.refresh_golden;
+            refresh_frame_flags.refresh_golden.Assign(refresh_frame_flags.refresh_alt.Value());
+            refresh_frame_flags.refresh_alt.Assign(temp);
         }
-
         if (current_frame_info.intra_only) {
             uncomp_writer.WriteU(frame_sync_code, 24);
-            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
+            uncomp_writer.WriteU(refresh_frame_flags.raw, 8);
             uncomp_writer.WriteU(current_frame_info.frame_size.width - 1, 16);
             uncomp_writer.WriteU(current_frame_info.frame_size.height - 1, 16);
             uncomp_writer.WriteBit(false); // Render and frame size different.
         } else {
-            uncomp_writer.WriteU(static_cast<s32>(refresh_frame_flags), 8);
-
-            for (s32 index = 1; index < 4; index++) {
+            const bool swap_indices = needs_ref_swap ^ swap_ref_indices;
+            const auto ref_frame_index = swap_indices ? std::array{0, 2, 1} : std::array{0, 1, 2};
+            uncomp_writer.WriteU(refresh_frame_flags.raw, 8);
+            for (size_t index = 1; index < 4; index++) {
                 uncomp_writer.WriteU(ref_frame_index[index - 1], 3);
                 uncomp_writer.WriteU(current_frame_info.ref_frame_sign_bias[index], 1);
             }
-
             uncomp_writer.WriteBit(true);  // Frame size with refs.
             uncomp_writer.WriteBit(false); // Render and frame size different.
             uncomp_writer.WriteBit(current_frame_info.allow_high_precision_mv);
@@ -714,10 +679,9 @@ VpxBitStreamWriter VP9::ComposeUncompressedHeader() {
         frame_ctx_idx = 1;
     }
 
-    uncomp_writer.WriteU(frame_ctx_idx, 2); // Frame context index.
-    prev_frame_probs =
-        frame_ctxs[frame_ctx_idx].probs; // reference probabilities for compressed header
-    frame_ctxs[frame_ctx_idx] = {current_frame_number, false, current_frame_info.entropy};
+    uncomp_writer.WriteU(frame_ctx_idx, 2);       // Frame context index.
+    prev_frame_probs = frame_ctxs[frame_ctx_idx]; // reference probabilities for compressed header
+    frame_ctxs[frame_ctx_idx] = current_frame_info.entropy;
 
     uncomp_writer.WriteU(current_frame_info.first_level, 6);
     uncomp_writer.WriteU(current_frame_info.sharpness_level, 3);
@@ -812,7 +776,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters
         current_frame_info = curr_frame.info;
         bitstream = std::move(curr_frame.bit_stream);
     }
-
     // The uncompressed header routine sets PrevProb parameters needed for the compressed header
     auto uncomp_writer = ComposeUncompressedHeader();
     std::vector<u8> compressed_header = ComposeCompressedHeader();
@@ -828,13 +791,6 @@ const std::vector<u8>& VP9::ComposeFrameHeader(const NvdecCommon::NvdecRegisters
               frame.begin() + uncompressed_header.size());
     std::copy(bitstream.begin(), bitstream.end(),
               frame.begin() + uncompressed_header.size() + compressed_header.size());
-
-    // keep track of frame number
-    current_frame_number++;
-    grace_period--;
-
-    // don't display hidden frames
-    hidden = !current_frame_info.show_frame;
     return frame;
 }
 
diff --git a/src/video_core/command_classes/codecs/vp9.h b/src/video_core/command_classes/codecs/vp9.h
index 8396c81054..e6e9fc17ef 100644
--- a/src/video_core/command_classes/codecs/vp9.h
+++ b/src/video_core/command_classes/codecs/vp9.h
@@ -14,7 +14,6 @@
 
 namespace Tegra {
 class GPU;
-enum class FrameType { KeyFrame = 0, InterFrame = 1 };
 namespace Decoder {
 
 /// The VpxRangeEncoder, and VpxBitStreamWriter classes are used to compose the
@@ -124,7 +123,7 @@ public:
 
     /// Returns true if the most recent frame was a hidden frame.
     [[nodiscard]] bool WasFrameHidden() const {
-        return hidden;
+        return !current_frame_info.show_frame;
     }
 
 private:
@@ -178,19 +177,12 @@ private:
     std::array<s8, 4> loop_filter_ref_deltas{};
     std::array<s8, 2> loop_filter_mode_deltas{};
 
-    bool hidden = false;
-    s64 current_frame_number = -2; // since we buffer 2 frames
-    s32 grace_period = 6;          // frame offsets need to stabilize
-    std::array<FrameContexts, 4> frame_ctxs{};
     Vp9FrameContainer next_frame{};
-    Vp9FrameContainer next_next_frame{};
-    bool swap_next_golden{};
+    std::array<Vp9EntropyProbs, 4> frame_ctxs{};
+    bool swap_ref_indices{};
 
     Vp9PictureInfo current_frame_info{};
     Vp9EntropyProbs prev_frame_probs{};
-
-    s32 diff_update_probability = 252;
-    s32 frame_sync_code = 0x498342;
 };
 
 } // namespace Decoder
diff --git a/src/video_core/command_classes/codecs/vp9_types.h b/src/video_core/command_classes/codecs/vp9_types.h
index 2da14f3ca5..6820afa266 100644
--- a/src/video_core/command_classes/codecs/vp9_types.h
+++ b/src/video_core/command_classes/codecs/vp9_types.h
@@ -296,12 +296,6 @@ struct RefPoolElement {
     bool refresh{};
 };
 
-struct FrameContexts {
-    s64 from;
-    bool adapted;
-    Vp9EntropyProbs probs;
-};
-
 #define ASSERT_POSITION(field_name, position)                                                      \
     static_assert(offsetof(Vp9EntropyProbs, field_name) == position,                               \
                   "Field " #field_name " has invalid position")
diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/command_classes/vic.cpp
index ffb7c82a1f..d5e77941cd 100644
--- a/src/video_core/command_classes/vic.cpp
+++ b/src/video_core/command_classes/vic.cpp
@@ -46,11 +46,8 @@ void Vic::ProcessMethod(Method method, u32 argument) {
     case Method::SetOutputSurfaceLumaOffset:
         output_surface_luma_address = arg;
         break;
-    case Method::SetOutputSurfaceChromaUOffset:
-        output_surface_chroma_u_address = arg;
-        break;
-    case Method::SetOutputSurfaceChromaVOffset:
-        output_surface_chroma_v_address = arg;
+    case Method::SetOutputSurfaceChromaOffset:
+        output_surface_chroma_address = arg;
         break;
     default:
         break;
@@ -65,11 +62,10 @@ void Vic::Execute() {
     const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)};
     const AVFramePtr frame_ptr = nvdec_processor->GetFrame();
     const auto* frame = frame_ptr.get();
-    if (!frame || frame->width == 0 || frame->height == 0) {
+    if (!frame) {
         return;
     }
-    const VideoPixelFormat pixel_format =
-        static_cast<VideoPixelFormat>(config.pixel_format.Value());
+    const auto pixel_format = static_cast<VideoPixelFormat>(config.pixel_format.Value());
     switch (pixel_format) {
     case VideoPixelFormat::BGRA8:
     case VideoPixelFormat::RGBA8: {
@@ -83,16 +79,18 @@ void Vic::Execute() {
             sws_freeContext(scaler_ctx);
             scaler_ctx = nullptr;
 
-            // FFmpeg returns all frames in YUV420, convert it into expected format
-            scaler_ctx =
-                sws_getContext(frame->width, frame->height, AV_PIX_FMT_YUV420P, frame->width,
-                               frame->height, target_format, 0, nullptr, nullptr, nullptr);
+            // Frames are decoded into either YUV420 or NV12 formats. Convert to desired format
+            scaler_ctx = sws_getContext(frame->width, frame->height,
+                                        static_cast<AVPixelFormat>(frame->format), frame->width,
+                                        frame->height, target_format, 0, nullptr, nullptr, nullptr);
 
             scaler_width = frame->width;
             scaler_height = frame->height;
         }
         // Get Converted frame
-        const std::size_t linear_size = frame->width * frame->height * 4;
+        const u32 width = static_cast<u32>(frame->width);
+        const u32 height = static_cast<u32>(frame->height);
+        const std::size_t linear_size = width * height * 4;
 
         // Only allocate frame_buffer once per stream, as the size is not expected to change
         if (!converted_frame_buffer) {
@@ -109,11 +107,10 @@ void Vic::Execute() {
         if (blk_kind != 0) {
             // swizzle pitch linear to block linear
             const u32 block_height = static_cast<u32>(config.block_linear_height_log2);
-            const auto size = Tegra::Texture::CalculateSize(true, 4, frame->width, frame->height, 1,
-                                                            block_height, 0);
+            const auto size =
+                Tegra::Texture::CalculateSize(true, 4, width, height, 1, block_height, 0);
             luma_buffer.resize(size);
-            Tegra::Texture::SwizzleSubrect(frame->width, frame->height, frame->width * 4,
-                                           frame->width, 4, luma_buffer.data(),
+            Tegra::Texture::SwizzleSubrect(width, height, width * 4, width, 4, luma_buffer.data(),
                                            converted_frame_buffer.get(), block_height, 0, 0);
 
             gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size);
@@ -131,41 +128,65 @@ void Vic::Execute() {
         const std::size_t surface_height = config.surface_height_minus1 + 1;
         const auto frame_width = std::min(surface_width, static_cast<size_t>(frame->width));
         const auto frame_height = std::min(surface_height, static_cast<size_t>(frame->height));
-        const std::size_t half_width = frame_width / 2;
-        const std::size_t half_height = frame_height / 2;
-        const std::size_t aligned_width = (surface_width + 0xff) & ~0xff;
+        const std::size_t aligned_width = (surface_width + 0xff) & ~0xffUL;
 
-        const auto* luma_ptr = frame->data[0];
-        const auto* chroma_b_ptr = frame->data[1];
-        const auto* chroma_r_ptr = frame->data[2];
         const auto stride = static_cast<size_t>(frame->linesize[0]);
-        const auto half_stride = static_cast<size_t>(frame->linesize[1]);
 
         luma_buffer.resize(aligned_width * surface_height);
         chroma_buffer.resize(aligned_width * surface_height / 2);
 
         // Populate luma buffer
+        const u8* luma_src = frame->data[0];
         for (std::size_t y = 0; y < frame_height; ++y) {
             const std::size_t src = y * stride;
             const std::size_t dst = y * aligned_width;
             for (std::size_t x = 0; x < frame_width; ++x) {
-                luma_buffer[dst + x] = luma_ptr[src + x];
+                luma_buffer[dst + x] = luma_src[src + x];
             }
         }
         gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(),
                                        luma_buffer.size());
 
-        // Populate chroma buffer from both channels with interleaving.
-        for (std::size_t y = 0; y < half_height; ++y) {
-            const std::size_t src = y * half_stride;
-            const std::size_t dst = y * aligned_width;
+        // Chroma
+        const std::size_t half_height = frame_height / 2;
+        const auto half_stride = static_cast<size_t>(frame->linesize[1]);
 
-            for (std::size_t x = 0; x < half_width; ++x) {
-                chroma_buffer[dst + x * 2] = chroma_b_ptr[src + x];
-                chroma_buffer[dst + x * 2 + 1] = chroma_r_ptr[src + x];
+        switch (frame->format) {
+        case AV_PIX_FMT_YUV420P: {
+            // Frame from FFmpeg software
+            // Populate chroma buffer from both channels with interleaving.
+            const std::size_t half_width = frame_width / 2;
+            const u8* chroma_b_src = frame->data[1];
+            const u8* chroma_r_src = frame->data[2];
+            for (std::size_t y = 0; y < half_height; ++y) {
+                const std::size_t src = y * half_stride;
+                const std::size_t dst = y * aligned_width;
+
+                for (std::size_t x = 0; x < half_width; ++x) {
+                    chroma_buffer[dst + x * 2] = chroma_b_src[src + x];
+                    chroma_buffer[dst + x * 2 + 1] = chroma_r_src[src + x];
+                }
+            }
+            break;
+        }
+        case AV_PIX_FMT_NV12: {
+            // Frame from VA-API hardware
+            // This is already interleaved so just copy
+            const u8* chroma_src = frame->data[1];
+            for (std::size_t y = 0; y < half_height; ++y) {
+                const std::size_t src = y * stride;
+                const std::size_t dst = y * aligned_width;
+                for (std::size_t x = 0; x < frame_width; ++x) {
+                    chroma_buffer[dst + x] = chroma_src[src + x];
+                }
             }
+            break;
+        }
+        default:
+            UNREACHABLE();
+            break;
         }
-        gpu.MemoryManager().WriteBlock(output_surface_chroma_u_address, chroma_buffer.data(),
+        gpu.MemoryManager().WriteBlock(output_surface_chroma_address, chroma_buffer.data(),
                                        chroma_buffer.size());
         break;
     }
diff --git a/src/video_core/command_classes/vic.h b/src/video_core/command_classes/vic.h
index f5a2ed100a..74246e08c9 100644
--- a/src/video_core/command_classes/vic.h
+++ b/src/video_core/command_classes/vic.h
@@ -22,8 +22,8 @@ public:
         SetControlParams = 0x1c1,
         SetConfigStructOffset = 0x1c2,
         SetOutputSurfaceLumaOffset = 0x1c8,
-        SetOutputSurfaceChromaUOffset = 0x1c9,
-        SetOutputSurfaceChromaVOffset = 0x1ca
+        SetOutputSurfaceChromaOffset = 0x1c9,
+        SetOutputSurfaceChromaUnusedOffset = 0x1ca
     };
 
     explicit Vic(GPU& gpu, std::shared_ptr<Nvdec> nvdec_processor);
@@ -64,8 +64,7 @@ private:
 
     GPUVAddr config_struct_address{};
     GPUVAddr output_surface_luma_address{};
-    GPUVAddr output_surface_chroma_u_address{};
-    GPUVAddr output_surface_chroma_v_address{};
+    GPUVAddr output_surface_chroma_address{};
 
     SwsContext* scaler_ctx{};
     s32 scaler_width{};
diff --git a/src/video_core/host_shaders/astc_decoder.comp b/src/video_core/host_shaders/astc_decoder.comp
index c37f15bfd8..f34c5f5d98 100644
--- a/src/video_core/host_shaders/astc_decoder.comp
+++ b/src/video_core/host_shaders/astc_decoder.comp
@@ -10,33 +10,27 @@
 #define END_PUSH_CONSTANTS };
 #define UNIFORM(n)
 #define BINDING_INPUT_BUFFER 0
-#define BINDING_ENC_BUFFER 1
-#define BINDING_SWIZZLE_BUFFER 2
-#define BINDING_OUTPUT_IMAGE 3
+#define BINDING_OUTPUT_IMAGE 1
 
 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
 
 #define BEGIN_PUSH_CONSTANTS
 #define END_PUSH_CONSTANTS
 #define UNIFORM(n) layout(location = n) uniform
-#define BINDING_SWIZZLE_BUFFER 0
-#define BINDING_INPUT_BUFFER 1
-#define BINDING_ENC_BUFFER 2
+#define BINDING_INPUT_BUFFER 0
 #define BINDING_OUTPUT_IMAGE 0
 
 #endif
 
-layout(local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
 
 BEGIN_PUSH_CONSTANTS
 UNIFORM(1) uvec2 block_dims;
-
-UNIFORM(2) uint bytes_per_block_log2;
-UNIFORM(3) uint layer_stride;
-UNIFORM(4) uint block_size;
-UNIFORM(5) uint x_shift;
-UNIFORM(6) uint block_height;
-UNIFORM(7) uint block_height_mask;
+UNIFORM(2) uint layer_stride;
+UNIFORM(3) uint block_size;
+UNIFORM(4) uint x_shift;
+UNIFORM(5) uint block_height;
+UNIFORM(6) uint block_height_mask;
 END_PUSH_CONSTANTS
 
 struct EncodingData {
@@ -55,45 +49,35 @@ struct TexelWeightParams {
     bool void_extent_hdr;
 };
 
-// Swizzle data
-layout(binding = BINDING_SWIZZLE_BUFFER, std430) readonly buffer SwizzleTable {
-    uint swizzle_table[];
-};
-
 layout(binding = BINDING_INPUT_BUFFER, std430) readonly buffer InputBufferU32 {
-    uint astc_data[];
-};
-
-// ASTC Encodings data
-layout(binding = BINDING_ENC_BUFFER, std430) readonly buffer EncodingsValues {
-    EncodingData encoding_values[];
+    uvec4 astc_data[];
 };
 
 layout(binding = BINDING_OUTPUT_IMAGE, rgba8) uniform writeonly image2DArray dest_image;
 
-const uint GOB_SIZE_X = 64;
-const uint GOB_SIZE_Y = 8;
-const uint GOB_SIZE_Z = 1;
-const uint GOB_SIZE = GOB_SIZE_X * GOB_SIZE_Y * GOB_SIZE_Z;
-
 const uint GOB_SIZE_X_SHIFT = 6;
 const uint GOB_SIZE_Y_SHIFT = 3;
-const uint GOB_SIZE_Z_SHIFT = 0;
-const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT + GOB_SIZE_Z_SHIFT;
-
-const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1, GOB_SIZE_Y - 1);
+const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
 
-const int BLOCK_SIZE_IN_BYTES = 16;
-
-const int BLOCK_INFO_ERROR = 0;
-const int BLOCK_INFO_VOID_EXTENT_HDR = 1;
-const int BLOCK_INFO_VOID_EXTENT_LDR = 2;
-const int BLOCK_INFO_NORMAL = 3;
+const uint BYTES_PER_BLOCK_LOG2 = 4;
 
 const int JUST_BITS = 0;
 const int QUINT = 1;
 const int TRIT = 2;
 
+// ASTC Encodings data, sorted in ascending order based on their BitLength value
+// (see GetBitLength() function)
+EncodingData encoding_values[22] = EncodingData[](
+    EncodingData(JUST_BITS, 0, 0, 0), EncodingData(JUST_BITS, 1, 0, 0), EncodingData(TRIT, 0, 0, 0),
+    EncodingData(JUST_BITS, 2, 0, 0), EncodingData(QUINT, 0, 0, 0), EncodingData(TRIT, 1, 0, 0),
+    EncodingData(JUST_BITS, 3, 0, 0), EncodingData(QUINT, 1, 0, 0), EncodingData(TRIT, 2, 0, 0),
+    EncodingData(JUST_BITS, 4, 0, 0), EncodingData(QUINT, 2, 0, 0), EncodingData(TRIT, 3, 0, 0),
+    EncodingData(JUST_BITS, 5, 0, 0), EncodingData(QUINT, 3, 0, 0), EncodingData(TRIT, 4, 0, 0),
+    EncodingData(JUST_BITS, 6, 0, 0), EncodingData(QUINT, 4, 0, 0), EncodingData(TRIT, 5, 0, 0),
+    EncodingData(JUST_BITS, 7, 0, 0), EncodingData(QUINT, 5, 0, 0), EncodingData(TRIT, 6, 0, 0),
+    EncodingData(JUST_BITS, 8, 0, 0)
+);
+
 // The following constants are expanded variants of the Replicate()
 // function calls corresponding to the following arguments:
 // value: index into the generated table
@@ -135,44 +119,37 @@ const uint REPLICATE_7_BIT_TO_8_TABLE[128] =
 // Input ASTC texture globals
 uint current_index = 0;
 int bitsread = 0;
-uint total_bitsread = 0;
-uint local_buff[16];
+int total_bitsread = 0;
+uvec4 local_buff;
 
 // Color data globals
-uint color_endpoint_data[16];
+uvec4 color_endpoint_data;
 int color_bitsread = 0;
-uint total_color_bitsread = 0;
-int color_index = 0;
 
 // Four values, two endpoints, four maximum paritions
 uint color_values[32];
 int colvals_index = 0;
 
 // Weight data globals
-uint texel_weight_data[16];
+uvec4 texel_weight_data;
 int texel_bitsread = 0;
-uint total_texel_bitsread = 0;
-int texel_index = 0;
 
 bool texel_flag = false;
 
 // Global "vectors" to be pushed into when decoding
-EncodingData result_vector[100];
+EncodingData result_vector[144];
 int result_index = 0;
 
-EncodingData texel_vector[100];
+EncodingData texel_vector[144];
 int texel_vector_index = 0;
 
 uint unquantized_texel_weights[2][144];
 
 uint SwizzleOffset(uvec2 pos) {
-    pos = pos & SWIZZLE_MASK;
-    return swizzle_table[pos.y * 64 + pos.x];
-}
-
-uint ReadTexel(uint offset) {
-    // extract the 8-bit value from the 32-bit packed data.
-    return bitfieldExtract(astc_data[offset / 4], int((offset * 8) & 24), 8);
+    uint x = pos.x;
+    uint y = pos.y;
+    return ((x % 64) / 32) * 256 + ((y % 8) / 2) * 64 + ((x % 32) / 16) * 32 +
+                          (y % 2) * 16 + (x % 16);
 }
 
 // Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
@@ -278,14 +255,10 @@ uint Hash52(uint p) {
     return p;
 }
 
-uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bool small_block) {
-    if (partition_count == 1) {
-        return 0;
-    }
+uint Select2DPartition(uint seed, uint x, uint y, uint partition_count, bool small_block) {
     if (small_block) {
         x <<= 1;
         y <<= 1;
-        z <<= 1;
     }
 
     seed += (partition_count - 1) * 1024;
@@ -299,10 +272,6 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo
     uint seed6 = uint((rnum >> 20) & 0xF);
     uint seed7 = uint((rnum >> 24) & 0xF);
     uint seed8 = uint((rnum >> 28) & 0xF);
-    uint seed9 = uint((rnum >> 18) & 0xF);
-    uint seed10 = uint((rnum >> 22) & 0xF);
-    uint seed11 = uint((rnum >> 26) & 0xF);
-    uint seed12 = uint(((rnum >> 30) | (rnum << 2)) & 0xF);
 
     seed1 = (seed1 * seed1);
     seed2 = (seed2 * seed2);
@@ -312,12 +281,8 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo
     seed6 = (seed6 * seed6);
     seed7 = (seed7 * seed7);
     seed8 = (seed8 * seed8);
-    seed9 = (seed9 * seed9);
-    seed10 = (seed10 * seed10);
-    seed11 = (seed11 * seed11);
-    seed12 = (seed12 * seed12);
 
-    int sh1, sh2, sh3;
+    uint sh1, sh2;
     if ((seed & 1) > 0) {
         sh1 = (seed & 2) > 0 ? 4 : 5;
         sh2 = (partition_count == 3) ? 6 : 5;
@@ -325,25 +290,19 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo
         sh1 = (partition_count == 3) ? 6 : 5;
         sh2 = (seed & 2) > 0 ? 4 : 5;
     }
-    sh3 = (seed & 0x10) > 0 ? sh1 : sh2;
-
-    seed1 = (seed1 >> sh1);
-    seed2 = (seed2 >> sh2);
-    seed3 = (seed3 >> sh1);
-    seed4 = (seed4 >> sh2);
-    seed5 = (seed5 >> sh1);
-    seed6 = (seed6 >> sh2);
-    seed7 = (seed7 >> sh1);
-    seed8 = (seed8 >> sh2);
-    seed9 = (seed9 >> sh3);
-    seed10 = (seed10 >> sh3);
-    seed11 = (seed11 >> sh3);
-    seed12 = (seed12 >> sh3);
-
-    uint a = seed1 * x + seed2 * y + seed11 * z + (rnum >> 14);
-    uint b = seed3 * x + seed4 * y + seed12 * z + (rnum >> 10);
-    uint c = seed5 * x + seed6 * y + seed9 * z + (rnum >> 6);
-    uint d = seed7 * x + seed8 * y + seed10 * z + (rnum >> 2);
+    seed1 >>= sh1;
+    seed2 >>= sh2;
+    seed3 >>= sh1;
+    seed4 >>= sh2;
+    seed5 >>= sh1;
+    seed6 >>= sh2;
+    seed7 >>= sh1;
+    seed8 >>= sh2;
+
+    uint a = seed1 * x + seed2 * y + (rnum >> 14);
+    uint b = seed3 * x + seed4 * y + (rnum >> 10);
+    uint c = seed5 * x + seed6 * y + (rnum >> 6);
+    uint d = seed7 * x + seed8 * y + (rnum >> 2);
 
     a &= 0x3F;
     b &= 0x3F;
@@ -368,58 +327,37 @@ uint SelectPartition(uint seed, uint x, uint y, uint z, uint partition_count, bo
     }
 }
 
-uint Select2DPartition(uint seed, uint x, uint y, uint partition_count, bool small_block) {
-    return SelectPartition(seed, x, y, 0, partition_count, small_block);
-}
-
-uint ReadBit() {
-    if (current_index >= local_buff.length()) {
+uint ExtractBits(uvec4 payload, int offset, int bits) {
+    if (bits <= 0) {
         return 0;
     }
-    uint bit = bitfieldExtract(local_buff[current_index], bitsread, 1);
-    ++bitsread;
-    ++total_bitsread;
-    if (bitsread == 8) {
-        ++current_index;
-        bitsread = 0;
+    int last_offset = offset + bits - 1;
+    int shifted_offset = offset >> 5;
+    if ((last_offset >> 5) == shifted_offset) {
+        return bitfieldExtract(payload[shifted_offset], offset & 31, bits);
     }
-    return bit;
+    int first_bits = 32 - (offset & 31);
+    int result_first = int(bitfieldExtract(payload[shifted_offset], offset & 31, first_bits));
+    int result_second = int(bitfieldExtract(payload[shifted_offset + 1], 0, bits - first_bits));
+    return result_first | (result_second << first_bits);
 }
 
 uint StreamBits(uint num_bits) {
-    uint ret = 0;
-    for (uint i = 0; i < num_bits; i++) {
-        ret |= ((ReadBit() & 1) << i);
-    }
+    int int_bits = int(num_bits);
+    uint ret = ExtractBits(local_buff, total_bitsread, int_bits);
+    total_bitsread += int_bits;
     return ret;
 }
 
-uint ReadColorBit() {
-    uint bit = 0;
-    if (texel_flag) {
-        bit = bitfieldExtract(texel_weight_data[texel_index], texel_bitsread, 1);
-        ++texel_bitsread;
-        ++total_texel_bitsread;
-        if (texel_bitsread == 8) {
-            ++texel_index;
-            texel_bitsread = 0;
-        }
-    } else {
-        bit = bitfieldExtract(color_endpoint_data[color_index], color_bitsread, 1);
-        ++color_bitsread;
-        ++total_color_bitsread;
-        if (color_bitsread == 8) {
-            ++color_index;
-            color_bitsread = 0;
-        }
-    }
-    return bit;
-}
-
 uint StreamColorBits(uint num_bits) {
     uint ret = 0;
-    for (uint i = 0; i < num_bits; i++) {
-        ret |= ((ReadColorBit() & 1) << i);
+    int int_bits = int(num_bits);
+    if (texel_flag) {
+        ret = ExtractBits(texel_weight_data, texel_bitsread, int_bits);
+        texel_bitsread += int_bits;
+    } else {
+        ret = ExtractBits(color_endpoint_data, color_bitsread, int_bits);
+        color_bitsread += int_bits;
     }
     return ret;
 }
@@ -596,22 +534,16 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits) {
     for (uint i = 0; i < num_partitions; i++) {
         num_values += ((modes[i] >> 2) + 1) << 1;
     }
-    int range = 256;
-    while (--range > 0) {
-        EncodingData val = encoding_values[range];
+    // Find the largest encoding that's within color_data_bits
+    // TODO(ameerj): profile with binary search
+    int range = 0;
+    while (++range < encoding_values.length()) {
         uint bit_length = GetBitLength(num_values, range);
-        if (bit_length <= color_data_bits) {
-            while (--range > 0) {
-                EncodingData newval = encoding_values[range];
-                if (newval.encoding != val.encoding && newval.num_bits != val.num_bits) {
-                    break;
-                }
-            }
-            ++range;
+        if (bit_length > color_data_bits) {
             break;
         }
     }
-    DecodeIntegerSequence(range, num_values);
+    DecodeIntegerSequence(range - 1, num_values);
     uint out_index = 0;
     for (int itr = 0; itr < result_index; ++itr) {
         if (out_index >= num_values) {
@@ -1028,7 +960,7 @@ int FindLayout(uint mode) {
     return 5;
 }
 
-TexelWeightParams DecodeBlockInfo(uint block_index) {
+TexelWeightParams DecodeBlockInfo() {
     TexelWeightParams params = TexelWeightParams(uvec2(0), 0, false, false, false, false);
     uint mode = StreamBits(11);
     if ((mode & 0x1ff) == 0x1fc) {
@@ -1110,10 +1042,10 @@ TexelWeightParams DecodeBlockInfo(uint block_index) {
     }
     weight_index -= 2;
     if ((mode_layout != 9) && ((mode & 0x200) != 0)) {
-        const int max_weights[6] = int[6](9, 11, 15, 19, 23, 31);
+        const int max_weights[6] = int[6](7, 8, 9, 10, 11, 12);
         params.max_weight = max_weights[weight_index];
     } else {
-        const int max_weights[6] = int[6](1, 2, 3, 4, 5, 7);
+        const int max_weights[6] = int[6](1, 2, 3, 4, 5, 6);
         params.max_weight = max_weights[weight_index];
     }
     return params;
@@ -1144,8 +1076,8 @@ void FillVoidExtentLDR(ivec3 coord) {
     }
 }
 
-void DecompressBlock(ivec3 coord, uint block_index) {
-    TexelWeightParams params = DecodeBlockInfo(block_index);
+void DecompressBlock(ivec3 coord) {
+    TexelWeightParams params = DecodeBlockInfo();
     if (params.error_state) {
         FillError(coord);
         return;
@@ -1212,7 +1144,7 @@ void DecompressBlock(ivec3 coord, uint block_index) {
     // Read color data...
     uint color_data_bits = remaining_bits;
     while (remaining_bits > 0) {
-        int nb = int(min(remaining_bits, 8U));
+        int nb = int(min(remaining_bits, 32U));
         uint b = StreamBits(nb);
         color_endpoint_data[ced_pointer] = uint(bitfieldExtract(b, 0, nb));
         ++ced_pointer;
@@ -1254,25 +1186,20 @@ void DecompressBlock(ivec3 coord, uint block_index) {
         ComputeEndpoints(endpoints[i][0], endpoints[i][1], color_endpoint_mode[i]);
     }
 
-    for (uint i = 0; i < 16; i++) {
-        texel_weight_data[i] = local_buff[i];
-    }
-    for (uint i = 0; i < 8; i++) {
-#define REVERSE_BYTE(b) ((b * 0x0802U & 0x22110U) | (b * 0x8020U & 0x88440U)) * 0x10101U >> 16
-        uint a = REVERSE_BYTE(texel_weight_data[i]);
-        uint b = REVERSE_BYTE(texel_weight_data[15 - i]);
-#undef REVERSE_BYTE
-        texel_weight_data[i] = uint(bitfieldExtract(b, 0, 8));
-        texel_weight_data[15 - i] = uint(bitfieldExtract(a, 0, 8));
-    }
+    texel_weight_data = local_buff;
+    texel_weight_data = bitfieldReverse(texel_weight_data).wzyx;
     uint clear_byte_start =
         (GetPackedBitSize(params.size, params.dual_plane, params.max_weight) >> 3) + 1;
-    texel_weight_data[clear_byte_start - 1] =
-        texel_weight_data[clear_byte_start - 1] &
+
+    uint byte_insert = ExtractBits(texel_weight_data, int(clear_byte_start - 1) * 8, 8) &
         uint(
             ((1 << (GetPackedBitSize(params.size, params.dual_plane, params.max_weight) % 8)) - 1));
-    for (uint i = 0; i < 16 - clear_byte_start; i++) {
-        texel_weight_data[clear_byte_start + i] = 0U;
+    uint vec_index = (clear_byte_start - 1) >> 2;
+    texel_weight_data[vec_index] =
+        bitfieldInsert(texel_weight_data[vec_index], byte_insert, int((clear_byte_start - 1) % 4) * 8, 8);
+    for (uint i = clear_byte_start; i < 16; ++i) {
+        uint idx = i >> 2;
+        texel_weight_data[idx] = bitfieldInsert(texel_weight_data[idx], 0, int(i % 4) * 8, 8);
     }
     texel_flag = true; // use texel "vector" and bit stream in integer decoding
     DecodeIntegerSequence(params.max_weight, GetNumWeightValues(params.size, params.dual_plane));
@@ -1281,8 +1208,11 @@ void DecompressBlock(ivec3 coord, uint block_index) {
 
     for (uint j = 0; j < block_dims.y; j++) {
         for (uint i = 0; i < block_dims.x; i++) {
-            uint local_partition = Select2DPartition(partition_index, i, j, num_partitions,
+            uint local_partition = 0;
+            if (num_partitions > 1) {
+                local_partition = Select2DPartition(partition_index, i, j, num_partitions,
                                                      (block_dims.y * block_dims.x) < 32);
+            }
             vec4 p;
             uvec4 C0 = ReplicateByteTo16(endpoints[local_partition][0]);
             uvec4 C1 = ReplicateByteTo16(endpoints[local_partition][1]);
@@ -1303,7 +1233,7 @@ void DecompressBlock(ivec3 coord, uint block_index) {
 
 void main() {
     uvec3 pos = gl_GlobalInvocationID;
-    pos.x <<= bytes_per_block_log2;
+    pos.x <<= BYTES_PER_BLOCK_LOG2;
 
     // Read as soon as possible due to its latency
     const uint swizzle = SwizzleOffset(pos.xy);
@@ -1321,13 +1251,8 @@ void main() {
     if (any(greaterThanEqual(coord, imageSize(dest_image)))) {
         return;
     }
-    uint block_index =
-        pos.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y + pos.y * gl_WorkGroupSize.x + pos.x;
-
     current_index = 0;
     bitsread = 0;
-    for (int i = 0; i < 16; i++) {
-        local_buff[i] = ReadTexel(offset + i);
-    }
-    DecompressBlock(coord, block_index);
+    local_buff = astc_data[offset / 16];
+    DecompressBlock(coord);
 }
diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp
index 37a4d1d9db..333f35a1c6 100644
--- a/src/video_core/renderer_opengl/util_shaders.cpp
+++ b/src/video_core/renderer_opengl/util_shaders.cpp
@@ -60,19 +60,14 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_)
       copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)) {
     const auto swizzle_table = Tegra::Texture::MakeSwizzleTable();
     swizzle_table_buffer.Create();
-    astc_buffer.Create();
     glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0);
-    glNamedBufferStorage(astc_buffer.handle, sizeof(ASTC_ENCODINGS_VALUES), &ASTC_ENCODINGS_VALUES,
-                         0);
 }
 
 UtilShaders::~UtilShaders() = default;
 
 void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map,
                              std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0;
-    static constexpr GLuint BINDING_INPUT_BUFFER = 1;
-    static constexpr GLuint BINDING_ENC_BUFFER = 2;
+    static constexpr GLuint BINDING_INPUT_BUFFER = 0;
     static constexpr GLuint BINDING_OUTPUT_IMAGE = 0;
 
     const Extent2D tile_size{
@@ -80,34 +75,32 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map,
         .height = VideoCore::Surface::DefaultBlockHeight(image.info.format),
     };
     program_manager.BindComputeProgram(astc_decoder_program.handle);
-    glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle);
-    glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_ENC_BUFFER, astc_buffer.handle);
-
     glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes);
     glUniform2ui(1, tile_size.width, tile_size.height);
+
     // Ensure buffer data is valid before dispatching
     glFlush();
     for (const SwizzleParameters& swizzle : swizzles) {
         const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
-        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U);
+        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 8U);
+        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U);
 
         const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);
         ASSERT(params.origin == (std::array<u32, 3>{0, 0, 0}));
         ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0}));
+        ASSERT(params.bytes_per_block_log2 == 4);
 
-        glUniform1ui(2, params.bytes_per_block_log2);
-        glUniform1ui(3, params.layer_stride);
-        glUniform1ui(4, params.block_size);
-        glUniform1ui(5, params.x_shift);
-        glUniform1ui(6, params.block_height);
-        glUniform1ui(7, params.block_height_mask);
+        glUniform1ui(2, params.layer_stride);
+        glUniform1ui(3, params.block_size);
+        glUniform1ui(4, params.x_shift);
+        glUniform1ui(5, params.block_height);
+        glUniform1ui(6, params.block_height_mask);
 
-        glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0,
-                           GL_WRITE_ONLY, GL_RGBA8);
         // ASTC texture data
         glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset,
                           image.guest_size_bytes - swizzle.buffer_offset);
+        glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0,
+                           GL_WRITE_ONLY, GL_RGBA8);
 
         glDispatchCompute(num_dispatches_x, num_dispatches_y, image.info.resources.layers);
     }
diff --git a/src/video_core/renderer_opengl/util_shaders.h b/src/video_core/renderer_opengl/util_shaders.h
index 53d65f368e..ef881e35f6 100644
--- a/src/video_core/renderer_opengl/util_shaders.h
+++ b/src/video_core/renderer_opengl/util_shaders.h
@@ -62,7 +62,6 @@ private:
     ProgramManager& program_manager;
 
     OGLBuffer swizzle_table_buffer;
-    OGLBuffer astc_buffer;
 
     OGLProgram astc_decoder_program;
     OGLProgram block_linear_unswizzle_2d_program;
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp
index 561cf5e11f..3e96c0f603 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp
@@ -30,16 +30,12 @@
 namespace Vulkan {
 
 using Tegra::Texture::SWIZZLE_TABLE;
-using Tegra::Texture::ASTC::ASTC_ENCODINGS_VALUES;
-using namespace Tegra::Texture::ASTC;
 
 namespace {
 
 constexpr u32 ASTC_BINDING_INPUT_BUFFER = 0;
-constexpr u32 ASTC_BINDING_ENC_BUFFER = 1;
-constexpr u32 ASTC_BINDING_SWIZZLE_BUFFER = 2;
-constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 3;
-constexpr size_t ASTC_NUM_BINDINGS = 4;
+constexpr u32 ASTC_BINDING_OUTPUT_IMAGE = 1;
+constexpr size_t ASTC_NUM_BINDINGS = 2;
 
 template <size_t size>
 inline constexpr VkPushConstantRange COMPUTE_PUSH_CONSTANT_RANGE{
@@ -75,7 +71,7 @@ constexpr DescriptorBankInfo INPUT_OUTPUT_BANK_INFO{
     .score = 2,
 };
 
-constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDINGS{{
+constexpr std::array<VkDescriptorSetLayoutBinding, ASTC_NUM_BINDINGS> ASTC_DESCRIPTOR_SET_BINDINGS{{
     {
         .binding = ASTC_BINDING_INPUT_BUFFER,
         .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
@@ -84,20 +80,6 @@ constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDIN
         .pImmutableSamplers = nullptr,
     },
     {
-        .binding = ASTC_BINDING_ENC_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-    {
-        .binding = ASTC_BINDING_SWIZZLE_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-    {
         .binding = ASTC_BINDING_OUTPUT_IMAGE,
         .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
         .descriptorCount = 1,
@@ -108,12 +90,12 @@ constexpr std::array<VkDescriptorSetLayoutBinding, 4> ASTC_DESCRIPTOR_SET_BINDIN
 
 constexpr DescriptorBankInfo ASTC_BANK_INFO{
     .uniform_buffers = 0,
-    .storage_buffers = 3,
+    .storage_buffers = 1,
     .texture_buffers = 0,
     .image_buffers = 0,
     .textures = 0,
     .images = 1,
-    .score = 4,
+    .score = 2,
 };
 
 constexpr VkDescriptorUpdateTemplateEntryKHR INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE{
@@ -136,22 +118,6 @@ constexpr std::array<VkDescriptorUpdateTemplateEntryKHR, ASTC_NUM_BINDINGS>
             .stride = sizeof(DescriptorUpdateEntry),
         },
         {
-            .dstBinding = ASTC_BINDING_ENC_BUFFER,
-            .dstArrayElement = 0,
-            .descriptorCount = 1,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-            .offset = ASTC_BINDING_ENC_BUFFER * sizeof(DescriptorUpdateEntry),
-            .stride = sizeof(DescriptorUpdateEntry),
-        },
-        {
-            .dstBinding = ASTC_BINDING_SWIZZLE_BUFFER,
-            .dstArrayElement = 0,
-            .descriptorCount = 1,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-            .offset = ASTC_BINDING_SWIZZLE_BUFFER * sizeof(DescriptorUpdateEntry),
-            .stride = sizeof(DescriptorUpdateEntry),
-        },
-        {
             .dstBinding = ASTC_BINDING_OUTPUT_IMAGE,
             .dstArrayElement = 0,
             .descriptorCount = 1,
@@ -163,7 +129,6 @@ constexpr std::array<VkDescriptorUpdateTemplateEntryKHR, ASTC_NUM_BINDINGS>
 
 struct AstcPushConstants {
     std::array<u32, 2> blocks_dims;
-    u32 bytes_per_block_log2;
     u32 layer_stride;
     u32 block_size;
     u32 x_shift;
@@ -354,46 +319,6 @@ ASTCDecoderPass::ASTCDecoderPass(const Device& device_, VKScheduler& scheduler_,
 
 ASTCDecoderPass::~ASTCDecoderPass() = default;
 
-void ASTCDecoderPass::MakeDataBuffer() {
-    constexpr size_t TOTAL_BUFFER_SIZE = sizeof(ASTC_ENCODINGS_VALUES) + sizeof(SWIZZLE_TABLE);
-    data_buffer = device.GetLogical().CreateBuffer(VkBufferCreateInfo{
-        .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = 0,
-        .size = TOTAL_BUFFER_SIZE,
-        .usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
-        .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
-        .queueFamilyIndexCount = 0,
-        .pQueueFamilyIndices = nullptr,
-    });
-    data_buffer_commit = memory_allocator.Commit(data_buffer, MemoryUsage::Upload);
-
-    const auto staging_ref = staging_buffer_pool.Request(TOTAL_BUFFER_SIZE, MemoryUsage::Upload);
-    std::memcpy(staging_ref.mapped_span.data(), &ASTC_ENCODINGS_VALUES,
-                sizeof(ASTC_ENCODINGS_VALUES));
-    // Tack on the swizzle table at the end of the buffer
-    std::memcpy(staging_ref.mapped_span.data() + sizeof(ASTC_ENCODINGS_VALUES), &SWIZZLE_TABLE,
-                sizeof(SWIZZLE_TABLE));
-
-    scheduler.Record([src = staging_ref.buffer, offset = staging_ref.offset, dst = *data_buffer,
-                      TOTAL_BUFFER_SIZE](vk::CommandBuffer cmdbuf) {
-        static constexpr VkMemoryBarrier write_barrier{
-            .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-        };
-        const VkBufferCopy copy{
-            .srcOffset = offset,
-            .dstOffset = 0,
-            .size = TOTAL_BUFFER_SIZE,
-        };
-        cmdbuf.CopyBuffer(src, dst, copy);
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
-                               0, write_barrier);
-    });
-}
-
 void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
                                std::span<const VideoCommon::SwizzleParameters> swizzles) {
     using namespace VideoCommon::Accelerated;
@@ -402,9 +327,6 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         VideoCore::Surface::DefaultBlockHeight(image.info.format),
     };
     scheduler.RequestOutsideRenderPassOperationContext();
-    if (!data_buffer) {
-        MakeDataBuffer();
-    }
     const VkPipeline vk_pipeline = *pipeline;
     const VkImageAspectFlags aspect_mask = image.AspectMask();
     const VkImage vk_image = image.Handle();
@@ -436,16 +358,13 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         });
     for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
         const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
-        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U);
+        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 8U);
+        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U);
         const u32 num_dispatches_z = image.info.resources.layers;
 
         update_descriptor_queue.Acquire();
         update_descriptor_queue.AddBuffer(map.buffer, input_offset,
                                           image.guest_size_bytes - swizzle.buffer_offset);
-        update_descriptor_queue.AddBuffer(*data_buffer, 0, sizeof(ASTC_ENCODINGS_VALUES));
-        update_descriptor_queue.AddBuffer(*data_buffer, sizeof(ASTC_ENCODINGS_VALUES),
-                                          sizeof(SWIZZLE_TABLE));
         update_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
         const void* const descriptor_data{update_descriptor_queue.UpdateData()};
 
@@ -453,11 +372,11 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);
         ASSERT(params.origin == (std::array<u32, 3>{0, 0, 0}));
         ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0}));
+        ASSERT(params.bytes_per_block_log2 == 4);
         scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_dispatches_z, block_dims,
                           params, descriptor_data](vk::CommandBuffer cmdbuf) {
             const AstcPushConstants uniforms{
                 .blocks_dims = block_dims,
-                .bytes_per_block_log2 = params.bytes_per_block_log2,
                 .layer_stride = params.layer_stride,
                 .block_size = params.block_size,
                 .x_shift = params.x_shift,
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h
index 114aef2bd7..c7b92cce0d 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pass.h
+++ b/src/video_core/renderer_vulkan/vk_compute_pass.h
@@ -96,15 +96,10 @@ public:
                   std::span<const VideoCommon::SwizzleParameters> swizzles);
 
 private:
-    void MakeDataBuffer();
-
     VKScheduler& scheduler;
     StagingBufferPool& staging_buffer_pool;
     VKUpdateDescriptorQueue& update_descriptor_queue;
     MemoryAllocator& memory_allocator;
-
-    vk::Buffer data_buffer;
-    MemoryCommit data_buffer_commit;
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index 3ab500760f..25161df1f6 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -151,6 +151,76 @@ private:
     const IntType& m_Bits;
 };
 
+enum class IntegerEncoding { JustBits, Quint, Trit };
+
+struct IntegerEncodedValue {
+    constexpr IntegerEncodedValue() = default;
+
+    constexpr IntegerEncodedValue(IntegerEncoding encoding_, u32 num_bits_)
+        : encoding{encoding_}, num_bits{num_bits_} {}
+
+    constexpr bool MatchesEncoding(const IntegerEncodedValue& other) const {
+        return encoding == other.encoding && num_bits == other.num_bits;
+    }
+
+    // Returns the number of bits required to encode num_vals values.
+    u32 GetBitLength(u32 num_vals) const {
+        u32 total_bits = num_bits * num_vals;
+        if (encoding == IntegerEncoding::Trit) {
+            total_bits += (num_vals * 8 + 4) / 5;
+        } else if (encoding == IntegerEncoding::Quint) {
+            total_bits += (num_vals * 7 + 2) / 3;
+        }
+        return total_bits;
+    }
+
+    IntegerEncoding encoding{};
+    u32 num_bits = 0;
+    u32 bit_value = 0;
+    union {
+        u32 quint_value = 0;
+        u32 trit_value;
+    };
+};
+
+// Returns a new instance of this struct that corresponds to the
+// can take no more than mav_value values
+static constexpr IntegerEncodedValue CreateEncoding(u32 mav_value) {
+    while (mav_value > 0) {
+        u32 check = mav_value + 1;
+
+        // Is mav_value a power of two?
+        if (!(check & (check - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::JustBits, std::popcount(mav_value));
+        }
+
+        // Is mav_value of the type 3*2^n - 1?
+        if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::Trit, std::popcount(check / 3 - 1));
+        }
+
+        // Is mav_value of the type 5*2^n - 1?
+        if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::Quint, std::popcount(check / 5 - 1));
+        }
+
+        // Apparently it can't be represented with a bounded integer sequence...
+        // just iterate.
+        mav_value--;
+    }
+    return IntegerEncodedValue(IntegerEncoding::JustBits, 0);
+}
+
+static constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() {
+    std::array<IntegerEncodedValue, 256> encodings{};
+    for (std::size_t i = 0; i < encodings.size(); ++i) {
+        encodings[i] = CreateEncoding(static_cast<u32>(i));
+    }
+    return encodings;
+}
+
+static constexpr std::array<IntegerEncodedValue, 256> ASTC_ENCODINGS_VALUES = MakeEncodedValues();
+
 namespace Tegra::Texture::ASTC {
 using IntegerEncodedVector = boost::container::static_vector<
     IntegerEncodedValue, 256,
@@ -521,35 +591,41 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     return params;
 }
 
-static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth,
-                              u32 blockHeight) {
-    // Don't actually care about the void extent, just read the bits...
-    for (s32 i = 0; i < 4; ++i) {
-        strm.ReadBits<13>();
+// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
+// is the same as [(num_bits - 1):0] and repeats all the way down.
+template <typename IntType>
+static constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) {
+    if (num_bits == 0 || to_bit == 0) {
+        return 0;
     }
-
-    // Decode the RGBA components and renormalize them to the range [0, 255]
-    u16 r = static_cast<u16>(strm.ReadBits<16>());
-    u16 g = static_cast<u16>(strm.ReadBits<16>());
-    u16 b = static_cast<u16>(strm.ReadBits<16>());
-    u16 a = static_cast<u16>(strm.ReadBits<16>());
-
-    u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 |
-               (static_cast<u32>(a) & 0xFF00) << 16;
-
-    for (u32 j = 0; j < blockHeight; j++) {
-        for (u32 i = 0; i < blockWidth; i++) {
-            outBuf[j * blockWidth + i] = rgba;
+    const IntType v = val & static_cast<IntType>((1 << num_bits) - 1);
+    IntType res = v;
+    u32 reslen = num_bits;
+    while (reslen < to_bit) {
+        u32 comp = 0;
+        if (num_bits > to_bit - reslen) {
+            u32 newshift = to_bit - reslen;
+            comp = num_bits - newshift;
+            num_bits = newshift;
         }
+        res = static_cast<IntType>(res << num_bits);
+        res = static_cast<IntType>(res | (v >> comp));
+        reslen += num_bits;
     }
+    return res;
 }
 
-static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
-    for (u32 j = 0; j < blockHeight; j++) {
-        for (u32 i = 0; i < blockWidth; i++) {
-            outBuf[j * blockWidth + i] = 0xFFFF00FF;
-        }
+static constexpr std::size_t NumReplicateEntries(u32 num_bits) {
+    return std::size_t(1) << num_bits;
+}
+
+template <typename IntType, u32 num_bits, u32 to_bit>
+static constexpr auto MakeReplicateTable() {
+    std::array<IntType, NumReplicateEntries(num_bits)> table{};
+    for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) {
+        table[value] = Replicate(value, num_bits, to_bit);
     }
+    return table;
 }
 
 static constexpr auto REPLICATE_BYTE_TO_16_TABLE = MakeReplicateTable<u32, 8, 16>();
@@ -572,6 +648,9 @@ static constexpr auto REPLICATE_2_BIT_TO_8_TABLE = MakeReplicateTable<u32, 2, 8>
 static constexpr auto REPLICATE_3_BIT_TO_8_TABLE = MakeReplicateTable<u32, 3, 8>();
 static constexpr auto REPLICATE_4_BIT_TO_8_TABLE = MakeReplicateTable<u32, 4, 8>();
 static constexpr auto REPLICATE_5_BIT_TO_8_TABLE = MakeReplicateTable<u32, 5, 8>();
+static constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>();
+static constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>();
+static constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>();
 /// Use a precompiled table with the most common usages, if it's not in the expected range, fallback
 /// to the runtime implementation
 static constexpr u32 FastReplicateTo8(u32 value, u32 num_bits) {
@@ -1316,6 +1395,37 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const u32*& colorValues,
 #undef READ_INT_VALUES
 }
 
+static void FillVoidExtentLDR(InputBitStream& strm, std::span<u32> outBuf, u32 blockWidth,
+                              u32 blockHeight) {
+    // Don't actually care about the void extent, just read the bits...
+    for (s32 i = 0; i < 4; ++i) {
+        strm.ReadBits<13>();
+    }
+
+    // Decode the RGBA components and renormalize them to the range [0, 255]
+    u16 r = static_cast<u16>(strm.ReadBits<16>());
+    u16 g = static_cast<u16>(strm.ReadBits<16>());
+    u16 b = static_cast<u16>(strm.ReadBits<16>());
+    u16 a = static_cast<u16>(strm.ReadBits<16>());
+
+    u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 |
+               (static_cast<u32>(a) & 0xFF00) << 16;
+
+    for (u32 j = 0; j < blockHeight; j++) {
+        for (u32 i = 0; i < blockWidth; i++) {
+            outBuf[j * blockWidth + i] = rgba;
+        }
+    }
+}
+
+static void FillError(std::span<u32> outBuf, u32 blockWidth, u32 blockHeight) {
+    for (u32 j = 0; j < blockHeight; j++) {
+        for (u32 i = 0; i < blockWidth; i++) {
+            outBuf[j * blockWidth + i] = 0xFFFF00FF;
+        }
+    }
+}
+
 static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth,
                             const u32 blockHeight, std::span<u32, 12 * 12> outBuf) {
     InputBitStream strm(inBuf);
diff --git a/src/video_core/textures/astc.h b/src/video_core/textures/astc.h
index 0229ae1220..14d2beec0c 100644
--- a/src/video_core/textures/astc.h
+++ b/src/video_core/textures/astc.h
@@ -9,117 +9,6 @@
 
 namespace Tegra::Texture::ASTC {
 
-enum class IntegerEncoding { JustBits, Quint, Trit };
-
-struct IntegerEncodedValue {
-    constexpr IntegerEncodedValue() = default;
-
-    constexpr IntegerEncodedValue(IntegerEncoding encoding_, u32 num_bits_)
-        : encoding{encoding_}, num_bits{num_bits_} {}
-
-    constexpr bool MatchesEncoding(const IntegerEncodedValue& other) const {
-        return encoding == other.encoding && num_bits == other.num_bits;
-    }
-
-    // Returns the number of bits required to encode num_vals values.
-    u32 GetBitLength(u32 num_vals) const {
-        u32 total_bits = num_bits * num_vals;
-        if (encoding == IntegerEncoding::Trit) {
-            total_bits += (num_vals * 8 + 4) / 5;
-        } else if (encoding == IntegerEncoding::Quint) {
-            total_bits += (num_vals * 7 + 2) / 3;
-        }
-        return total_bits;
-    }
-
-    IntegerEncoding encoding{};
-    u32 num_bits = 0;
-    u32 bit_value = 0;
-    union {
-        u32 quint_value = 0;
-        u32 trit_value;
-    };
-};
-
-// Returns a new instance of this struct that corresponds to the
-// can take no more than mav_value values
-constexpr IntegerEncodedValue CreateEncoding(u32 mav_value) {
-    while (mav_value > 0) {
-        u32 check = mav_value + 1;
-
-        // Is mav_value a power of two?
-        if (!(check & (check - 1))) {
-            return IntegerEncodedValue(IntegerEncoding::JustBits, std::popcount(mav_value));
-        }
-
-        // Is mav_value of the type 3*2^n - 1?
-        if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) {
-            return IntegerEncodedValue(IntegerEncoding::Trit, std::popcount(check / 3 - 1));
-        }
-
-        // Is mav_value of the type 5*2^n - 1?
-        if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) {
-            return IntegerEncodedValue(IntegerEncoding::Quint, std::popcount(check / 5 - 1));
-        }
-
-        // Apparently it can't be represented with a bounded integer sequence...
-        // just iterate.
-        mav_value--;
-    }
-    return IntegerEncodedValue(IntegerEncoding::JustBits, 0);
-}
-
-constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() {
-    std::array<IntegerEncodedValue, 256> encodings{};
-    for (std::size_t i = 0; i < encodings.size(); ++i) {
-        encodings[i] = CreateEncoding(static_cast<u32>(i));
-    }
-    return encodings;
-}
-
-constexpr std::array<IntegerEncodedValue, 256> ASTC_ENCODINGS_VALUES = MakeEncodedValues();
-
-// Replicates low num_bits such that [(to_bit - 1):(to_bit - 1 - from_bit)]
-// is the same as [(num_bits - 1):0] and repeats all the way down.
-template <typename IntType>
-constexpr IntType Replicate(IntType val, u32 num_bits, u32 to_bit) {
-    if (num_bits == 0 || to_bit == 0) {
-        return 0;
-    }
-    const IntType v = val & static_cast<IntType>((1 << num_bits) - 1);
-    IntType res = v;
-    u32 reslen = num_bits;
-    while (reslen < to_bit) {
-        u32 comp = 0;
-        if (num_bits > to_bit - reslen) {
-            u32 newshift = to_bit - reslen;
-            comp = num_bits - newshift;
-            num_bits = newshift;
-        }
-        res = static_cast<IntType>(res << num_bits);
-        res = static_cast<IntType>(res | (v >> comp));
-        reslen += num_bits;
-    }
-    return res;
-}
-
-constexpr std::size_t NumReplicateEntries(u32 num_bits) {
-    return std::size_t(1) << num_bits;
-}
-
-template <typename IntType, u32 num_bits, u32 to_bit>
-constexpr auto MakeReplicateTable() {
-    std::array<IntType, NumReplicateEntries(num_bits)> table{};
-    for (IntType value = 0; value < static_cast<IntType>(std::size(table)); ++value) {
-        table[value] = Replicate(value, num_bits, to_bit);
-    }
-    return table;
-}
-
-constexpr auto REPLICATE_6_BIT_TO_8_TABLE = MakeReplicateTable<u32, 6, 8>();
-constexpr auto REPLICATE_7_BIT_TO_8_TABLE = MakeReplicateTable<u32, 7, 8>();
-constexpr auto REPLICATE_8_BIT_TO_8_TABLE = MakeReplicateTable<u32, 8, 8>();
-
 void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
                 uint32_t block_width, uint32_t block_height, std::span<uint8_t> output);
 
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index f1f523ad10..c32ae956a3 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -18,9 +18,9 @@
 
 namespace Tegra::Texture {
 namespace {
-template <bool TO_LINEAR>
-void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
-             u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) {
+template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
+void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth,
+                 u32 block_height, u32 block_depth, u32 stride_alignment) {
     // The origin of the transformation can be configured here, leave it as zero as the current API
     // doesn't expose it.
     static constexpr u32 origin_x = 0;
@@ -28,9 +28,9 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe
     static constexpr u32 origin_z = 0;
 
     // We can configure here a custom pitch
-    // As it's not exposed 'width * bpp' will be the expected pitch.
-    const u32 pitch = width * bytes_per_pixel;
-    const u32 stride = Common::AlignUpLog2(width, stride_alignment) * bytes_per_pixel;
+    // As it's not exposed 'width * BYTES_PER_PIXEL' will be the expected pitch.
+    const u32 pitch = width * BYTES_PER_PIXEL;
+    const u32 stride = Common::AlignUpLog2(width, stride_alignment) * BYTES_PER_PIXEL;
 
     const u32 gobs_in_x = Common::DivCeilLog2(stride, GOB_SIZE_X_SHIFT);
     const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
@@ -54,14 +54,14 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe
                                  ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
 
             for (u32 column = 0; column < width; ++column) {
-                const u32 x = (column + origin_x) * bytes_per_pixel;
+                const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
                 const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
 
                 const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
                 const u32 swizzled_offset = base_swizzled_offset + table[x % GOB_SIZE_X];
 
                 const u32 unswizzled_offset =
-                    slice * pitch * height + line * pitch + column * bytes_per_pixel;
+                    slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
 
                 if (const auto offset = (TO_LINEAR ? unswizzled_offset : swizzled_offset);
                     offset >= input.size()) {
@@ -73,11 +73,45 @@ void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixe
 
                 u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
                 const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
-                std::memcpy(dst, src, bytes_per_pixel);
+
+                std::memcpy(dst, src, BYTES_PER_PIXEL);
             }
         }
     }
 }
+
+template <bool TO_LINEAR>
+void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
+             u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) {
+    switch (bytes_per_pixel) {
+    case 1:
+        return SwizzleImpl<TO_LINEAR, 1>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 2:
+        return SwizzleImpl<TO_LINEAR, 2>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 3:
+        return SwizzleImpl<TO_LINEAR, 3>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 4:
+        return SwizzleImpl<TO_LINEAR, 4>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 6:
+        return SwizzleImpl<TO_LINEAR, 6>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 8:
+        return SwizzleImpl<TO_LINEAR, 8>(output, input, width, height, depth, block_height,
+                                         block_depth, stride_alignment);
+    case 12:
+        return SwizzleImpl<TO_LINEAR, 12>(output, input, width, height, depth, block_height,
+                                          block_depth, stride_alignment);
+    case 16:
+        return SwizzleImpl<TO_LINEAR, 16>(output, input, width, height, depth, block_height,
+                                          block_depth, stride_alignment);
+    default:
+        UNREACHABLE_MSG("Invalid bytes_per_pixel={}", bytes_per_pixel);
+    }
+}
 } // Anonymous namespace
 
 void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 2b20fca8a9..380379eb4f 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -873,10 +873,6 @@ void Config::ReadShortcutValues() {
 void Config::ReadSystemValues() {
     qt_config->beginGroup(QStringLiteral("System"));
 
-    ReadBasicSetting(Settings::values.current_user);
-    Settings::values.current_user = std::clamp<int>(Settings::values.current_user.GetValue(), 0,
-                                                    Service::Account::MAX_USERS - 1);
-
     ReadGlobalSetting(Settings::values.language_index);
 
     ReadGlobalSetting(Settings::values.region_index);
@@ -897,6 +893,10 @@ void Config::ReadSystemValues() {
     }
 
     if (global) {
+        ReadBasicSetting(Settings::values.current_user);
+        Settings::values.current_user = std::clamp<int>(Settings::values.current_user.GetValue(), 0,
+                                                        Service::Account::MAX_USERS - 1);
+
         const auto custom_rtc_enabled =
             ReadSetting(QStringLiteral("custom_rtc_enabled"), false).toBool();
         if (custom_rtc_enabled) {
@@ -1406,7 +1406,6 @@ void Config::SaveShortcutValues() {
 void Config::SaveSystemValues() {
     qt_config->beginGroup(QStringLiteral("System"));
 
-    WriteBasicSetting(Settings::values.current_user);
     WriteGlobalSetting(Settings::values.language_index);
     WriteGlobalSetting(Settings::values.region_index);
     WriteGlobalSetting(Settings::values.time_zone_index);
@@ -1418,6 +1417,8 @@ void Config::SaveSystemValues() {
                  0, Settings::values.rng_seed.UsingGlobal());
 
     if (global) {
+        WriteBasicSetting(Settings::values.current_user);
+
         WriteSetting(QStringLiteral("custom_rtc_enabled"), Settings::values.custom_rtc.has_value(),
                      false);
         WriteSetting(QStringLiteral("custom_rtc"),
diff --git a/src/yuzu/configuration/configure_input_player_widget.cpp b/src/yuzu/configuration/configure_input_player_widget.cpp
index f50cda2f3c..cd633e45fb 100644
--- a/src/yuzu/configuration/configure_input_player_widget.cpp
+++ b/src/yuzu/configuration/configure_input_player_widget.cpp
@@ -122,6 +122,7 @@ void PlayerControlPreview::UpdateColors() {
         colors.slider_arrow = QColor(14, 15, 18);
         colors.font2 = QColor(255, 255, 255);
         colors.indicator = QColor(170, 238, 255);
+        colors.indicator2 = QColor(100, 255, 100);
         colors.deadzone = QColor(204, 136, 136);
         colors.slider_button = colors.button;
     }
@@ -139,6 +140,7 @@ void PlayerControlPreview::UpdateColors() {
         colors.slider_arrow = QColor(65, 68, 73);
         colors.font2 = QColor(0, 0, 0);
         colors.indicator = QColor(0, 0, 200);
+        colors.indicator2 = QColor(0, 150, 0);
         colors.deadzone = QColor(170, 0, 0);
         colors.slider_button = QColor(153, 149, 149);
     }
@@ -317,8 +319,7 @@ void PlayerControlPreview::DrawLeftController(QPainter& p, const QPointF center)
         using namespace Settings::NativeAnalog;
         DrawJoystick(p, center + QPointF(9, -69) + (axis_values[LStick].value * 8), 1.8f,
                      button_values[Settings::NativeButton::LStick]);
-        DrawRawJoystick(p, center + QPointF(-140, 90), axis_values[LStick].raw_value,
-                        axis_values[LStick].properties);
+        DrawRawJoystick(p, center + QPointF(-140, 90), QPointF(0, 0));
     }
 
     using namespace Settings::NativeButton;
@@ -432,8 +433,7 @@ void PlayerControlPreview::DrawRightController(QPainter& p, const QPointF center
         using namespace Settings::NativeAnalog;
         DrawJoystick(p, center + QPointF(-9, 11) + (axis_values[RStick].value * 8), 1.8f,
                      button_values[Settings::NativeButton::RStick]);
-        DrawRawJoystick(p, center + QPointF(140, 90), axis_values[RStick].raw_value,
-                        axis_values[RStick].properties);
+        DrawRawJoystick(p, QPointF(0, 0), center + QPointF(140, 90));
     }
 
     using namespace Settings::NativeButton;
@@ -547,8 +547,7 @@ void PlayerControlPreview::DrawDualController(QPainter& p, const QPointF center)
 
         DrawJoystick(p, center + QPointF(-65, -65) + (l_stick.value * 7), 1.62f, l_button);
         DrawJoystick(p, center + QPointF(65, 12) + (r_stick.value * 7), 1.62f, r_button);
-        DrawRawJoystick(p, center + QPointF(-180, 90), l_stick.raw_value, l_stick.properties);
-        DrawRawJoystick(p, center + QPointF(180, 90), r_stick.raw_value, r_stick.properties);
+        DrawRawJoystick(p, center + QPointF(-180, 90), center + QPointF(180, 90));
     }
 
     using namespace Settings::NativeButton;
@@ -634,8 +633,7 @@ void PlayerControlPreview::DrawHandheldController(QPainter& p, const QPointF cen
 
         DrawJoystick(p, center + QPointF(-171, -41) + (l_stick.value * 4), 1.0f, l_button);
         DrawJoystick(p, center + QPointF(171, 8) + (r_stick.value * 4), 1.0f, r_button);
-        DrawRawJoystick(p, center + QPointF(-50, 0), l_stick.raw_value, l_stick.properties);
-        DrawRawJoystick(p, center + QPointF(50, 0), r_stick.raw_value, r_stick.properties);
+        DrawRawJoystick(p, center + QPointF(-50, 0), center + QPointF(50, 0));
     }
 
     using namespace Settings::NativeButton;
@@ -728,10 +726,7 @@ void PlayerControlPreview::DrawProController(QPainter& p, const QPointF center)
                         button_values[Settings::NativeButton::LStick]);
         DrawProJoystick(p, center + QPointF(51, 0), axis_values[RStick].value, 11,
                         button_values[Settings::NativeButton::RStick]);
-        DrawRawJoystick(p, center + QPointF(-50, 105), axis_values[LStick].raw_value,
-                        axis_values[LStick].properties);
-        DrawRawJoystick(p, center + QPointF(50, 105), axis_values[RStick].raw_value,
-                        axis_values[RStick].properties);
+        DrawRawJoystick(p, center + QPointF(-50, 105), center + QPointF(50, 105));
     }
 
     using namespace Settings::NativeButton;
@@ -821,10 +816,7 @@ void PlayerControlPreview::DrawGCController(QPainter& p, const QPointF center) {
         p.setBrush(colors.font);
         DrawSymbol(p, center + QPointF(61, 37) + (axis_values[RStick].value * 9.5f), Symbol::C,
                    1.0f);
-        DrawRawJoystick(p, center + QPointF(-198, -125), axis_values[LStick].raw_value,
-                        axis_values[LStick].properties);
-        DrawRawJoystick(p, center + QPointF(198, -125), axis_values[RStick].raw_value,
-                        axis_values[RStick].properties);
+        DrawRawJoystick(p, center + QPointF(-198, -125), center + QPointF(198, -125));
     }
 
     using namespace Settings::NativeButton;
@@ -2358,8 +2350,33 @@ void PlayerControlPreview::DrawGCJoystick(QPainter& p, const QPointF center, boo
     DrawCircle(p, center, 7.5f);
 }
 
-void PlayerControlPreview::DrawRawJoystick(QPainter& p, const QPointF center, const QPointF value,
-                                           const Input::AnalogProperties& properties) {
+void PlayerControlPreview::DrawRawJoystick(QPainter& p, QPointF center_left, QPointF center_right) {
+    using namespace Settings::NativeAnalog;
+    if (controller_type != Settings::ControllerType::LeftJoycon) {
+        DrawJoystickProperties(p, center_right, axis_values[RStick].properties);
+        p.setPen(colors.indicator);
+        p.setBrush(colors.indicator);
+        DrawJoystickDot(p, center_right, axis_values[RStick].raw_value,
+                        axis_values[RStick].properties);
+        p.setPen(colors.indicator2);
+        p.setBrush(colors.indicator2);
+        DrawJoystickDot(p, center_right, axis_values[RStick].value, axis_values[RStick].properties);
+    }
+
+    if (controller_type != Settings::ControllerType::RightJoycon) {
+        DrawJoystickProperties(p, center_left, axis_values[LStick].properties);
+        p.setPen(colors.indicator);
+        p.setBrush(colors.indicator);
+        DrawJoystickDot(p, center_left, axis_values[LStick].raw_value,
+                        axis_values[LStick].properties);
+        p.setPen(colors.indicator2);
+        p.setBrush(colors.indicator2);
+        DrawJoystickDot(p, center_left, axis_values[LStick].value, axis_values[LStick].properties);
+    }
+}
+
+void PlayerControlPreview::DrawJoystickProperties(QPainter& p, const QPointF center,
+                                                  const Input::AnalogProperties& properties) {
     constexpr float size = 45.0f;
     const float range = size * properties.range;
     const float deadzone = size * properties.deadzone;
@@ -2376,10 +2393,14 @@ void PlayerControlPreview::DrawRawJoystick(QPainter& p, const QPointF center, co
     pen.setColor(colors.deadzone);
     p.setPen(pen);
     DrawCircle(p, center, deadzone);
+}
+
+void PlayerControlPreview::DrawJoystickDot(QPainter& p, const QPointF center, const QPointF value,
+                                           const Input::AnalogProperties& properties) {
+    constexpr float size = 45.0f;
+    const float range = size * properties.range;
 
     // Dot pointer
-    p.setPen(colors.indicator);
-    p.setBrush(colors.indicator);
     DrawCircle(p, center + (value * range), 2);
 }
 
diff --git a/src/yuzu/configuration/configure_input_player_widget.h b/src/yuzu/configuration/configure_input_player_widget.h
index 5fc16d8af0..f4a6a5e1b2 100644
--- a/src/yuzu/configuration/configure_input_player_widget.h
+++ b/src/yuzu/configuration/configure_input_player_widget.h
@@ -90,6 +90,7 @@ private:
         QColor highlight2{};
         QColor transparent{};
         QColor indicator{};
+        QColor indicator2{};
         QColor led_on{};
         QColor led_off{};
         QColor slider{};
@@ -139,7 +140,10 @@ private:
     // Draw joystick functions
     void DrawJoystick(QPainter& p, QPointF center, float size, bool pressed);
     void DrawJoystickSideview(QPainter& p, QPointF center, float angle, float size, bool pressed);
-    void DrawRawJoystick(QPainter& p, QPointF center, QPointF value,
+    void DrawRawJoystick(QPainter& p, QPointF center_left, QPointF center_right);
+    void DrawJoystickProperties(QPainter& p, QPointF center,
+                                const Input::AnalogProperties& properties);
+    void DrawJoystickDot(QPainter& p, QPointF center, QPointF value,
                          const Input::AnalogProperties& properties);
     void DrawProJoystick(QPainter& p, QPointF center, QPointF offset, float scalar, bool pressed);
     void DrawGCJoystick(QPainter& p, QPointF center, bool pressed);
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 064ecaafa3..4f14be524c 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -278,6 +278,9 @@ void Config::ReadValues() {
             if (Settings::values.players.GetValue()[p].analogs[i].empty())
                 Settings::values.players.GetValue()[p].analogs[i] = default_param;
         }
+
+        Settings::values.players.GetValue()[p].connected =
+            sdl2_config->GetBoolean(group, "connected", false);
     }
 
     ReadSetting("ControlsGeneral", Settings::values.mouse_enabled);