vfs_concat: fix time complexity of read

Add a 5ms tiemout to the DSP processing wait

README.md

@@ -83,5 +83,3 @@ If you wish to support us a different way, please join our [Discord](https://dis
 ## License
 
 yuzu is licensed under the GPLv3 (or any later version). Refer to the [LICENSE.txt](https://github.com/yuzu-emu/yuzu/blob/master/LICENSE.txt) file.
-
-The [Skyline-Emulator Team](https://github.com/skyline-emu/skyline) may choose to use the code from these contributors under the GPL-3.0-or-later OR MPL-2.0: [FernandoS27](https://github.com/FernandoS27), [lioncash](https://github.com/lioncash), [bunnei](https://github.com/bunnei), [ReinUsesLisp](https://github.com/ReinUsesLisp), [Morph1984](https://github.com/Morph1984), [ogniK5377](https://github.com/ogniK5377), [german77](https://github.com/german77), [ameerj](https://github.com/ameerj), [Kelebek1](https://github.com/Kelebek1) and [lat9nq](https://github.com/lat9nq)

externals/CMakeLists.txt

@@ -139,3 +139,7 @@ if (NOT TARGET LLVM::Demangle)
     target_sources(demangle PRIVATE demangle/ItaniumDemangle.cpp)
     add_library(LLVM::Demangle ALIAS demangle)
 endif()
+
+add_library(stb STATIC)
+target_include_directories(stb PUBLIC ./stb)
+target_sources(stb PRIVATE stb/stb_dxt.cpp)

externals/stb/stb_dxt.cpp

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

externals/stb/stb_dxt.h

@@ -0,0 +1,36 @@
+// SPDX-FileCopyrightText: fabian "ryg" giesen
+// SPDX-License-Identifier: MIT
+
+// stb_dxt.h - v1.12 - DXT1/DXT5 compressor
+
+#ifndef STB_INCLUDE_STB_DXT_H
+#define STB_INCLUDE_STB_DXT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef STB_DXT_STATIC
+#define STBDDEF static
+#else
+#define STBDDEF extern
+#endif
+
+// compression mode (bitflags)
+#define STB_DXT_NORMAL 0
+#define STB_DXT_DITHER 1 // use dithering. was always dubious, now deprecated. does nothing!
+#define STB_DXT_HIGHQUAL                                                                           \
+    2 // high quality mode, does two refinement steps instead of 1. ~30-40% slower.
+
+STBDDEF void stb_compress_bc1_block(unsigned char* dest,
+                                    const unsigned char* src_rgba_four_bytes_per_pixel, int alpha,
+                                    int mode);
+
+STBDDEF void stb_compress_bc3_block(unsigned char* dest, const unsigned char* src, int mode);
+
+#define STB_COMPRESS_DXT_BLOCK
+
+#ifdef __cplusplus
+}
+#endif
+#endif // STB_INCLUDE_STB_DXT_H

src/audio_core/renderer/adsp/audio_renderer.cpp

@@ -154,6 +154,11 @@ void AudioRenderer::ThreadFunc() {
             return;
 
         case RenderMessage::AudioRenderer_Render: {
+            if (system.IsShuttingDown()) [[unlikely]] {
+                std::this_thread::sleep_for(std::chrono::milliseconds(5));
+                mailbox->ADSPSendMessage(RenderMessage::AudioRenderer_RenderResponse);
+                continue;
+            }
             std::array<bool, MaxRendererSessions> buffers_reset{};
             std::array<u64, MaxRendererSessions> render_times_taken{};
             const auto start_time{system.CoreTiming().GetClockTicks()};

src/audio_core/renderer/system_manager.cpp

@@ -27,7 +27,7 @@ bool SystemManager::InitializeUnsafe() {
     if (!active) {
         if (adsp.Start()) {
             active = true;
-            thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(); });
+            thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(stop_token); });
         }
     }
 
@@ -39,8 +39,7 @@ void SystemManager::Stop() {
         return;
     }
     active = false;
-    update.store(true);
-    update.notify_all();
+    thread.request_stop();
     thread.join();
     adsp.Stop();
 }
@@ -85,12 +84,12 @@ bool SystemManager::Remove(System& system_) {
     return true;
 }
 
-void SystemManager::ThreadFunc() {
+void SystemManager::ThreadFunc(std::stop_token stop_token) {
     static constexpr char name[]{"AudioRenderSystemManager"};
     MicroProfileOnThreadCreate(name);
     Common::SetCurrentThreadName(name);
     Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
-    while (active) {
+    while (active && !stop_token.stop_requested()) {
         {
             std::scoped_lock l{mutex1};
 

src/audio_core/renderer/system_manager.h

@@ -66,13 +66,7 @@ private:
     /**
      * Main thread responsible for command generation.
      */
-    void ThreadFunc();
-
-    enum class StreamState {
-        Filling,
-        Steady,
-        Draining,
-    };
+    void ThreadFunc(std::stop_token stop_token);
 
     /// Core system
     Core::System& core;
@@ -90,8 +84,6 @@ private:
     ADSP::ADSP& adsp;
     /// AudioRenderer mailbox for communication
     ADSP::AudioRenderer_Mailbox* mailbox{};
-    /// Atomic for main thread to wait on
-    std::atomic<bool> update{};
 };
 
 } // namespace AudioCore::AudioRenderer

src/audio_core/sink/sink_stream.cpp

@@ -271,8 +271,8 @@ u64 SinkStream::GetExpectedPlayedSampleCount() {
 
 void SinkStream::WaitFreeSpace() {
     std::unique_lock lk{release_mutex};
-    release_cv.wait(
-        lk, [this]() { return queued_buffers < max_queue_size || system.IsShuttingDown(); });
+    release_cv.wait_for(lk, std::chrono::milliseconds(5),
+                        [this]() { return queued_buffers < max_queue_size; });
 }
 
 } // namespace AudioCore::Sink

src/common/scratch_buffer.h

@@ -23,7 +23,10 @@ public:
           buffer{Common::make_unique_for_overwrite<T[]>(initial_capacity)} {}
 
     ~ScratchBuffer() = default;
+    ScratchBuffer(const ScratchBuffer&) = delete;
+    ScratchBuffer& operator=(const ScratchBuffer&) = delete;
     ScratchBuffer(ScratchBuffer&&) = default;
+    ScratchBuffer& operator=(ScratchBuffer&&) = default;
 
     /// This will only grow the buffer's capacity if size is greater than the current capacity.
     /// The previously held data will remain intact.
@@ -87,6 +90,12 @@ public:
         return buffer_capacity;
     }
 
+    void swap(ScratchBuffer& other) noexcept {
+        std::swap(last_requested_size, other.last_requested_size);
+        std::swap(buffer_capacity, other.buffer_capacity);
+        std::swap(buffer, other.buffer);
+    }
+
 private:
     size_t last_requested_size{};
     size_t buffer_capacity{};

src/common/settings.cpp

@@ -61,7 +61,9 @@ void LogSettings() {
     log_setting("Renderer_NvdecEmulation", values.nvdec_emulation.GetValue());
     log_setting("Renderer_AccelerateASTC", values.accelerate_astc.GetValue());
     log_setting("Renderer_AsyncASTC", values.async_astc.GetValue());
+    log_setting("Renderer_AstcRecompression", values.astc_recompression.GetValue());
     log_setting("Renderer_UseVsync", values.vsync_mode.GetValue());
+    log_setting("Renderer_UseReactiveFlushing", values.use_reactive_flushing.GetValue());
     log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
     log_setting("Renderer_UseAsynchronousShaders", values.use_asynchronous_shaders.GetValue());
     log_setting("Renderer_AnisotropicFilteringLevel", values.max_anisotropy.GetValue());
@@ -223,6 +225,8 @@ void RestoreGlobalState(bool is_powered_on) {
     values.nvdec_emulation.SetGlobal(true);
     values.accelerate_astc.SetGlobal(true);
     values.async_astc.SetGlobal(true);
+    values.astc_recompression.SetGlobal(true);
+    values.use_reactive_flushing.SetGlobal(true);
     values.shader_backend.SetGlobal(true);
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);
@@ -230,6 +234,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.bg_red.SetGlobal(true);
     values.bg_green.SetGlobal(true);
     values.bg_blue.SetGlobal(true);
+    values.enable_compute_pipelines.SetGlobal(true);
 
     // System
     values.language_index.SetGlobal(true);

src/common/settings.h

@@ -90,6 +90,12 @@ enum class AntiAliasing : u32 {
     LastAA = Smaa,
 };
 
+enum class AstcRecompression : u32 {
+    Uncompressed = 0,
+    Bc1 = 1,
+    Bc3 = 2,
+};
+
 struct ResolutionScalingInfo {
     u32 up_scale{1};
     u32 down_shift{0};
@@ -465,12 +471,17 @@ struct Values {
     SwitchableSetting<bool> async_astc{false, "async_astc"};
     Setting<VSyncMode, true> vsync_mode{VSyncMode::FIFO, VSyncMode::Immediate,
                                         VSyncMode::FIFORelaxed, "use_vsync"};
+    SwitchableSetting<bool> use_reactive_flushing{true, "use_reactive_flushing"};
     SwitchableSetting<ShaderBackend, true> shader_backend{ShaderBackend::GLSL, ShaderBackend::GLSL,
                                                           ShaderBackend::SPIRV, "shader_backend"};
     SwitchableSetting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};
     SwitchableSetting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
     SwitchableSetting<bool> use_vulkan_driver_pipeline_cache{true,
                                                              "use_vulkan_driver_pipeline_cache"};
+    SwitchableSetting<bool> enable_compute_pipelines{false, "enable_compute_pipelines"};
+    SwitchableSetting<AstcRecompression, true> astc_recompression{
+        AstcRecompression::Uncompressed, AstcRecompression::Uncompressed, AstcRecompression::Bc3,
+        "astc_recompression"};
 
     SwitchableSetting<u8> bg_red{0, "bg_red"};
     SwitchableSetting<u8> bg_green{0, "bg_green"};
@@ -534,6 +545,8 @@ struct Values {
     Setting<bool> enable_ir_sensor{false, "enable_ir_sensor"};
     Setting<std::string> ir_sensor_device{"auto", "ir_sensor_device"};
 
+    Setting<bool> random_amiibo_id{false, "random_amiibo_id"};
+
     // Data Storage
     Setting<bool> use_virtual_sd{true, "use_virtual_sd"};
     Setting<bool> gamecard_inserted{false, "gamecard_inserted"};

src/core/core.cpp

@@ -117,8 +117,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
             return nullptr;
         }
 
-        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(std::move(concat),
-                                                                  dir->GetName());
+        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(concat, dir->GetName());
     }
 
     if (Common::FS::IsDir(path)) {
@@ -612,6 +611,10 @@ void System::PrepareReschedule(const u32 core_index) {
     impl->kernel.PrepareReschedule(core_index);
 }
 
+size_t System::GetCurrentHostThreadID() const {
+    return impl->kernel.GetCurrentHostThreadID();
+}
+
 PerfStatsResults System::GetAndResetPerfStats() {
     return impl->GetAndResetPerfStats();
 }

src/core/core.h

@@ -222,6 +222,8 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
+    [[nodiscard]] size_t GetCurrentHostThreadID() const;
+
     /// Gets and resets core performance statistics
     [[nodiscard]] PerfStatsResults GetAndResetPerfStats();
 

src/core/file_sys/romfs.cpp

@@ -140,7 +140,8 @@ VirtualFile CreateRomFS(VirtualDir dir, VirtualDir ext) {
         return nullptr;
 
     RomFSBuildContext ctx{dir, ext};
-    return ConcatenatedVfsFile::MakeConcatenatedFile(0, ctx.Build(), dir->GetName());
+    auto file_map = ctx.Build();
+    return ConcatenatedVfsFile::MakeConcatenatedFile(0, file_map, dir->GetName());
 }
 
 } // namespace FileSys

src/core/file_sys/savedata_factory.cpp

@@ -82,9 +82,9 @@ std::string GetFutureSaveDataPath(SaveDataSpaceId space_id, SaveDataType type, u
     // Only detect account/device saves from the future location.
     switch (type) {
     case SaveDataType::SaveData:
-        return fmt::format("{}/account/{}/{:016X}/1", space_id_path, uuid.RawString(), title_id);
+        return fmt::format("{}/account/{}/{:016X}/0", space_id_path, uuid.RawString(), title_id);
     case SaveDataType::DeviceSaveData:
-        return fmt::format("{}/device/{:016X}/1", space_id_path, title_id);
+        return fmt::format("{}/device/{:016X}/0", space_id_path, title_id);
     default:
         return "";
     }

src/core/file_sys/vfs_concat.cpp

@@ -10,84 +10,105 @@
 
 namespace FileSys {
 
-static bool VerifyConcatenationMapContinuity(const std::multimap<u64, VirtualFile>& map) {
-    const auto last_valid = --map.end();
-    for (auto iter = map.begin(); iter != last_valid;) {
-        const auto old = iter++;
-        if (old->first + old->second->GetSize() != iter->first) {
+ConcatenatedVfsFile::ConcatenatedVfsFile(ConcatenationMap&& concatenation_map_, std::string&& name_)
+    : concatenation_map(std::move(concatenation_map_)), name(std::move(name_)) {
+    DEBUG_ASSERT(this->VerifyContinuity());
+}
+
+bool ConcatenatedVfsFile::VerifyContinuity() const {
+    u64 last_offset = 0;
+    for (auto& entry : concatenation_map) {
+        if (entry.offset != last_offset) {
             return false;
         }
+
+        last_offset = entry.offset + entry.file->GetSize();
     }
 
-    return map.begin()->first == 0;
-}
-
-ConcatenatedVfsFile::ConcatenatedVfsFile(std::vector<VirtualFile> files_, std::string name_)
-    : name(std::move(name_)) {
-    std::size_t next_offset = 0;
-    for (const auto& file : files_) {
-        files.emplace(next_offset, file);
-        next_offset += file->GetSize();
-    }
-}
-
-ConcatenatedVfsFile::ConcatenatedVfsFile(std::multimap<u64, VirtualFile> files_, std::string name_)
-    : files(std::move(files_)), name(std::move(name_)) {
-    ASSERT(VerifyConcatenationMapContinuity(files));
+    return true;
 }
 
 ConcatenatedVfsFile::~ConcatenatedVfsFile() = default;
 
-VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(std::vector<VirtualFile> files,
-                                                      std::string name) {
-    if (files.empty())
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(const std::vector<VirtualFile>& files,
+                                                      std::string&& name) {
+    // Fold trivial cases.
+    if (files.empty()) {
         return nullptr;
-    if (files.size() == 1)
-        return files[0];
+    }
+    if (files.size() == 1) {
+        return files.front();
+    }
 
-    return VirtualFile(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+    // Make the concatenation map from the input.
+    std::vector<ConcatenationEntry> concatenation_map;
+    concatenation_map.reserve(files.size());
+    u64 last_offset = 0;
+
+    for (auto& file : files) {
+        concatenation_map.emplace_back(ConcatenationEntry{
+            .offset = last_offset,
+            .file = file,
+        });
+
+        last_offset += file->GetSize();
+    }
+
+    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
 }
 
 VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
-                                                      std::multimap<u64, VirtualFile> files,
-                                                      std::string name) {
-    if (files.empty())
+                                                      const std::multimap<u64, VirtualFile>& files,
+                                                      std::string&& name) {
+    // Fold trivial cases.
+    if (files.empty()) {
         return nullptr;
-    if (files.size() == 1)
+    }
+    if (files.size() == 1) {
         return files.begin()->second;
-
-    const auto last_valid = --files.end();
-    for (auto iter = files.begin(); iter != last_valid;) {
-        const auto old = iter++;
-        if (old->first + old->second->GetSize() != iter->first) {
-            files.emplace(old->first + old->second->GetSize(),
-                          std::make_shared<StaticVfsFile>(filler_byte, iter->first - old->first -
-                                                                           old->second->GetSize()));
-        }
     }
 
-    // Ensure the map starts at offset 0 (start of file), otherwise pad to fill.
-    if (files.begin()->first != 0)
-        files.emplace(0, std::make_shared<StaticVfsFile>(filler_byte, files.begin()->first));
+    // Make the concatenation map from the input.
+    std::vector<ConcatenationEntry> concatenation_map;
 
-    return VirtualFile(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+    concatenation_map.reserve(files.size());
+    u64 last_offset = 0;
+
+    // Iteration of a multimap is ordered, so offset will be strictly non-decreasing.
+    for (auto& [offset, file] : files) {
+        if (offset > last_offset) {
+            concatenation_map.emplace_back(ConcatenationEntry{
+                .offset = last_offset,
+                .file = std::make_shared<StaticVfsFile>(filler_byte, offset - last_offset),
+            });
+        }
+
+        concatenation_map.emplace_back(ConcatenationEntry{
+            .offset = offset,
+            .file = file,
+        });
+
+        last_offset = offset + file->GetSize();
+    }
+
+    return VirtualFile(new ConcatenatedVfsFile(std::move(concatenation_map), std::move(name)));
 }
 
 std::string ConcatenatedVfsFile::GetName() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
         return "";
     }
     if (!name.empty()) {
         return name;
     }
-    return files.begin()->second->GetName();
+    return concatenation_map.front().file->GetName();
 }
 
 std::size_t ConcatenatedVfsFile::GetSize() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
         return 0;
     }
-    return files.rbegin()->first + files.rbegin()->second->GetSize();
+    return concatenation_map.back().offset + concatenation_map.back().file->GetSize();
 }
 
 bool ConcatenatedVfsFile::Resize(std::size_t new_size) {
@@ -95,10 +116,10 @@ bool ConcatenatedVfsFile::Resize(std::size_t new_size) {
 }
 
 VirtualDir ConcatenatedVfsFile::GetContainingDirectory() const {
-    if (files.empty()) {
+    if (concatenation_map.empty()) {
         return nullptr;
     }
-    return files.begin()->second->GetContainingDirectory();
+    return concatenation_map.front().file->GetContainingDirectory();
 }
 
 bool ConcatenatedVfsFile::IsWritable() const {
@@ -110,25 +131,45 @@ bool ConcatenatedVfsFile::IsReadable() const {
 }
 
 std::size_t ConcatenatedVfsFile::Read(u8* data, std::size_t length, std::size_t offset) const {
-    auto entry = --files.end();
-    for (auto iter = files.begin(); iter != files.end(); ++iter) {
-        if (iter->first > offset) {
-            entry = --iter;
+    const ConcatenationEntry key{
+        .offset = offset,
+        .file = nullptr,
+    };
+
+    // Read nothing if the map is empty.
+    if (concatenation_map.empty()) {
+        return 0;
+    }
+
+    // Binary search to find the iterator to the first position we can check.
+    // It must exist, since we are not empty and are comparing unsigned integers.
+    auto it = std::prev(std::upper_bound(concatenation_map.begin(), concatenation_map.end(), key));
+    u64 cur_length = length;
+    u64 cur_offset = offset;
+
+    while (cur_length > 0 && it != concatenation_map.end()) {
+        // Check if we can read the file at this position.
+        const auto& file = it->file;
+        const u64 file_offset = it->offset;
+        const u64 file_size = file->GetSize();
+
+        if (cur_offset >= file_offset + file_size) {
+            // Entirely out of bounds read.
             break;
         }
+
+        // Read the file at this position.
+        const u64 intended_read_size = std::min<u64>(cur_length, file_size);
+        const u64 actual_read_size =
+            file->Read(data + (cur_offset - offset), intended_read_size, cur_offset - file_offset);
+
+        // Update tracking.
+        cur_offset += actual_read_size;
+        cur_length -= actual_read_size;
+        it++;
     }
 
-    if (entry->first + entry->second->GetSize() <= offset)
-        return 0;
-
-    const auto read_in =
-        std::min<u64>(entry->first + entry->second->GetSize() - offset, entry->second->GetSize());
-    if (length > read_in) {
-        return entry->second->Read(data, read_in, offset - entry->first) +
-               Read(data + read_in, length - read_in, offset + read_in);
-    }
-
-    return entry->second->Read(data, std::min<u64>(read_in, length), offset - entry->first);
+    return cur_offset - offset;
 }
 
 std::size_t ConcatenatedVfsFile::Write(const u8* data, std::size_t length, std::size_t offset) {

src/core/file_sys/vfs_concat.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <compare>
 #include <map>
 #include <memory>
 #include "core/file_sys/vfs.h"
@@ -12,19 +13,33 @@ namespace FileSys {
 // Class that wraps multiple vfs files and concatenates them, making reads seamless. Currently
 // read-only.
 class ConcatenatedVfsFile : public VfsFile {
-    explicit ConcatenatedVfsFile(std::vector<VirtualFile> files, std::string name_);
-    explicit ConcatenatedVfsFile(std::multimap<u64, VirtualFile> files, std::string name_);
+private:
+    struct ConcatenationEntry {
+        u64 offset;
+        VirtualFile file;
+
+        auto operator<=>(const ConcatenationEntry& other) const {
+            return this->offset <=> other.offset;
+        }
+    };
+    using ConcatenationMap = std::vector<ConcatenationEntry>;
+
+    explicit ConcatenatedVfsFile(std::vector<ConcatenationEntry>&& concatenation_map,
+                                 std::string&& name);
+    bool VerifyContinuity() const;
 
 public:
     ~ConcatenatedVfsFile() override;
 
     /// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
-    static VirtualFile MakeConcatenatedFile(std::vector<VirtualFile> files, std::string name);
+    static VirtualFile MakeConcatenatedFile(const std::vector<VirtualFile>& files,
+                                            std::string&& name);
 
     /// Convenience function that turns a map of offsets to files into a concatenated file, filling
     /// gaps with a given filler byte.
-    static VirtualFile MakeConcatenatedFile(u8 filler_byte, std::multimap<u64, VirtualFile> files,
-                                            std::string name);
+    static VirtualFile MakeConcatenatedFile(u8 filler_byte,
+                                            const std::multimap<u64, VirtualFile>& files,
+                                            std::string&& name);
 
     std::string GetName() const override;
     std::size_t GetSize() const override;
@@ -37,8 +52,7 @@ public:
     bool Rename(std::string_view new_name) override;
 
 private:
-    // Maps starting offset to file -- more efficient.
-    std::multimap<u64, VirtualFile> files;
+    ConcatenationMap concatenation_map;
     std::string name;
 };
 

src/core/file_sys/vfs_layered.cpp

@@ -2,6 +2,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <algorithm>
+#include <set>
 #include <utility>
 #include "core/file_sys/vfs_layered.h"
 
@@ -58,11 +59,13 @@ std::string LayeredVfsDirectory::GetFullPath() const {
 
 std::vector<VirtualFile> LayeredVfsDirectory::GetFiles() const {
     std::vector<VirtualFile> out;
+    std::set<std::string, std::less<>> out_names;
+
     for (const auto& layer : dirs) {
         for (const auto& file : layer->GetFiles()) {
-            if (std::find_if(out.begin(), out.end(), [&file](const VirtualFile& comp) {
-                    return comp->GetName() == file->GetName();
-                }) == out.end()) {
+            auto file_name = file->GetName();
+            if (!out_names.contains(file_name)) {
+                out_names.emplace(std::move(file_name));
                 out.push_back(file);
             }
         }

src/core/file_sys/vfs_vector.cpp

@@ -67,6 +67,23 @@ VectorVfsDirectory::VectorVfsDirectory(std::vector<VirtualFile> files_,
 
 VectorVfsDirectory::~VectorVfsDirectory() = default;
 
+VirtualFile VectorVfsDirectory::GetFile(std::string_view file_name) const {
+    if (!optimized_file_index_built) {
+        optimized_file_index.clear();
+        for (size_t i = 0; i < files.size(); i++) {
+            optimized_file_index.emplace(files[i]->GetName(), i);
+        }
+        optimized_file_index_built = true;
+    }
+
+    const auto it = optimized_file_index.find(file_name);
+    if (it != optimized_file_index.end()) {
+        return files[it->second];
+    }
+
+    return nullptr;
+}
+
 std::vector<VirtualFile> VectorVfsDirectory::GetFiles() const {
     return files;
 }
@@ -107,6 +124,7 @@ bool VectorVfsDirectory::DeleteSubdirectory(std::string_view subdir_name) {
 }
 
 bool VectorVfsDirectory::DeleteFile(std::string_view file_name) {
+    optimized_file_index_built = false;
     return FindAndRemoveVectorElement(files, file_name);
 }
 
@@ -124,6 +142,7 @@ VirtualFile VectorVfsDirectory::CreateFile(std::string_view file_name) {
 }
 
 void VectorVfsDirectory::AddFile(VirtualFile file) {
+    optimized_file_index_built = false;
     files.push_back(std::move(file));
 }
 

src/core/file_sys/vfs_vector.h

@@ -105,6 +105,7 @@ public:
                                 VirtualDir parent = nullptr);
     ~VectorVfsDirectory() override;
 
+    VirtualFile GetFile(std::string_view file_name) const override;
     std::vector<VirtualFile> GetFiles() const override;
     std::vector<VirtualDir> GetSubdirectories() const override;
     bool IsWritable() const override;
@@ -126,6 +127,9 @@ private:
 
     VirtualDir parent;
     std::string name;
+
+    mutable std::map<std::string, size_t, std::less<>> optimized_file_index;
+    mutable bool optimized_file_index_built{};
 };
 
 } // namespace FileSys

src/core/hid/emulated_controller.cpp

@@ -688,6 +688,12 @@ void EmulatedController::SetMotionParam(std::size_t index, Common::ParamPackage
     ReloadInput();
 }
 
+void EmulatedController::StartMotionCalibration() {
+    for (ControllerMotionInfo& motion : controller.motion_values) {
+        motion.emulated.Calibrate();
+    }
+}
+
 void EmulatedController::SetButton(const Common::Input::CallbackStatus& callback, std::size_t index,
                                    Common::UUID uuid) {
     if (index >= controller.button_values.size()) {
@@ -979,7 +985,6 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
     emulated.SetUserGyroThreshold(raw_status.gyro.x.properties.threshold);
     emulated.UpdateRotation(raw_status.delta_timestamp);
     emulated.UpdateOrientation(raw_status.delta_timestamp);
-    force_update_motion = raw_status.force_update;
 
     auto& motion = controller.motion_state[index];
     motion.accel = emulated.GetAcceleration();
@@ -1278,9 +1283,14 @@ bool EmulatedController::HasNfc() const {
 }
 
 bool EmulatedController::WriteNfc(const std::vector<u8>& data) {
-    auto& nfc_output_device = output_devices[3];
+    auto& nfc_output_device = output_devices[static_cast<std::size_t>(DeviceIndex::Right)];
+    auto& nfc_virtual_output_device = output_devices[3];
 
-    return nfc_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
+    if (nfc_output_device->SupportsNfc() != Common::Input::NfcState::NotSupported) {
+        return nfc_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
+    }
+
+    return nfc_virtual_output_device->WriteNfcData(data) == Common::Input::NfcState::Success;
 }
 
 void EmulatedController::SetLedPattern() {
@@ -1618,19 +1628,6 @@ NpadGcTriggerState EmulatedController::GetTriggers() const {
 
 MotionState EmulatedController::GetMotions() const {
     std::unique_lock lock{mutex};
-
-    // Some drivers like mouse motion need constant refreshing
-    if (force_update_motion) {
-        for (auto& device : motion_devices) {
-            if (!device) {
-                continue;
-            }
-            lock.unlock();
-            device->ForceUpdate();
-            lock.lock();
-        }
-    }
-
     return controller.motion_state;
 }
 
@@ -1696,8 +1693,21 @@ void EmulatedController::DeleteCallback(int key) {
     callback_list.erase(iterator);
 }
 
-void EmulatedController::TurboButtonUpdate() {
+void EmulatedController::StatusUpdate() {
     turbo_button_state = (turbo_button_state + 1) % (TURBO_BUTTON_DELAY * 2);
+
+    // Some drivers like key motion need constant refreshing
+    for (std::size_t index = 0; index < motion_devices.size(); ++index) {
+        const auto& raw_status = controller.motion_values[index].raw_status;
+        auto& device = motion_devices[index];
+        if (!raw_status.force_update) {
+            continue;
+        }
+        if (!device) {
+            continue;
+        }
+        device->ForceUpdate();
+    }
 }
 
 NpadButton EmulatedController::GetTurboButtonMask() const {

src/core/hid/emulated_controller.h

@@ -290,6 +290,9 @@ public:
      */
     void SetMotionParam(std::size_t index, Common::ParamPackage param);
 
+    /// Auto calibrates the current motion devices
+    void StartMotionCalibration();
+
     /// Returns the latest button status from the controller with parameters
     ButtonValues GetButtonsValues() const;
 
@@ -415,8 +418,8 @@ public:
      */
     void DeleteCallback(int key);
 
-    /// Swaps the state of the turbo buttons
-    void TurboButtonUpdate();
+    /// Swaps the state of the turbo buttons and updates motion input
+    void StatusUpdate();
 
 private:
     /// creates input devices from params
@@ -528,7 +531,6 @@ private:
     bool is_configuring{false};
     bool system_buttons_enabled{true};
     f32 motion_sensitivity{Core::HID::MotionInput::IsAtRestStandard};
-    bool force_update_motion{false};
     u32 turbo_button_state{0};
 
     // Temporary values to avoid doing changes while the controller is in configuring mode

src/core/hid/input_converter.cpp

@@ -86,7 +86,7 @@ Common::Input::MotionStatus TransformToMotion(const Common::Input::CallbackStatu
             .range = 1.0f,
             .offset = 0.0f,
         };
-        status.delta_timestamp = 5000;
+        status.delta_timestamp = 1000;
         status.force_update = true;
         status.accel.x = {
             .value = 0.0f,

src/core/hid/motion_input.cpp

@@ -37,11 +37,17 @@ void MotionInput::SetGyroscope(const Common::Vec3f& gyroscope) {
     gyro.y = std::clamp(gyro.y, -GyroMaxValue, GyroMaxValue);
     gyro.z = std::clamp(gyro.z, -GyroMaxValue, GyroMaxValue);
 
-    // Auto adjust drift to minimize drift
+    // Auto adjust gyro_bias to minimize drift
     if (!IsMoving(IsAtRestRelaxed)) {
         gyro_bias = (gyro_bias * 0.9999f) + (gyroscope * 0.0001f);
     }
 
+    // Adjust drift when calibration mode is enabled
+    if (calibration_mode) {
+        gyro_bias = (gyro_bias * 0.99f) + (gyroscope * 0.01f);
+        StopCalibration();
+    }
+
     if (gyro.Length() < gyro_threshold * user_gyro_threshold) {
         gyro = {};
     } else {
@@ -107,6 +113,19 @@ void MotionInput::UpdateRotation(u64 elapsed_time) {
     rotations += gyro * sample_period;
 }
 
+void MotionInput::Calibrate() {
+    calibration_mode = true;
+    calibration_counter = 0;
+}
+
+void MotionInput::StopCalibration() {
+    if (calibration_counter++ > CalibrationSamples) {
+        calibration_mode = false;
+        ResetQuaternion();
+        ResetRotations();
+    }
+}
+
 // Based on Madgwick's implementation of Mayhony's AHRS algorithm.
 // https://github.com/xioTechnologies/Open-Source-AHRS-With-x-IMU/blob/master/x-IMU%20IMU%20and%20AHRS%20Algorithms/x-IMU%20IMU%20and%20AHRS%20Algorithms/AHRS/MahonyAHRS.cs
 void MotionInput::UpdateOrientation(u64 elapsed_time) {

src/core/hid/motion_input.h

@@ -23,6 +23,8 @@ public:
     static constexpr float GyroMaxValue = 5.0f;
     static constexpr float AccelMaxValue = 7.0f;
 
+    static constexpr std::size_t CalibrationSamples = 300;
+
     explicit MotionInput();
 
     MotionInput(const MotionInput&) = default;
@@ -49,6 +51,8 @@ public:
     void UpdateRotation(u64 elapsed_time);
     void UpdateOrientation(u64 elapsed_time);
 
+    void Calibrate();
+
     [[nodiscard]] std::array<Common::Vec3f, 3> GetOrientation() const;
     [[nodiscard]] Common::Vec3f GetAcceleration() const;
     [[nodiscard]] Common::Vec3f GetGyroscope() const;
@@ -61,6 +65,7 @@ public:
     [[nodiscard]] bool IsCalibrated(f32 sensitivity) const;
 
 private:
+    void StopCalibration();
     void ResetOrientation();
     void SetOrientationFromAccelerometer();
 
@@ -103,6 +108,12 @@ private:
 
     // Use accelerometer values to calculate position
     bool only_accelerometer = true;
+
+    // When enabled it will aggressively adjust for gyro drift
+    bool calibration_mode = false;
+
+    // Used to auto disable calibration mode
+    std::size_t calibration_counter = 0;
 };
 
 } // namespace Core::HID

src/core/hle/kernel/k_memory_block_manager.h

@@ -144,14 +144,10 @@ private:
 
 class KScopedMemoryBlockManagerAuditor {
 public:
-    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager* m) : m_manager(m) {
-        ASSERT(m_manager->CheckState());
-    }
+    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager* m) : m_manager(m) {}
     explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager& m)
         : KScopedMemoryBlockManagerAuditor(std::addressof(m)) {}
-    ~KScopedMemoryBlockManagerAuditor() {
-        ASSERT(m_manager->CheckState());
-    }
+    ~KScopedMemoryBlockManagerAuditor() = default;
 
 private:
     KMemoryBlockManager* m_manager;

src/core/hle/service/am/am.cpp

@@ -13,6 +13,7 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_transfer_memory.h"
+#include "core/hle/result.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applet_ae.h"
@@ -1335,7 +1336,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {24, nullptr, "GetLaunchStorageInfoForDebug"},
         {25, &IApplicationFunctions::ExtendSaveData, "ExtendSaveData"},
         {26, &IApplicationFunctions::GetSaveDataSize, "GetSaveDataSize"},
-        {27, nullptr, "CreateCacheStorage"},
+        {27, &IApplicationFunctions::CreateCacheStorage, "CreateCacheStorage"},
         {28, nullptr, "GetSaveDataSizeMax"},
         {29, nullptr, "GetCacheStorageMax"},
         {30, &IApplicationFunctions::BeginBlockingHomeButtonShortAndLongPressed, "BeginBlockingHomeButtonShortAndLongPressed"},
@@ -1738,6 +1739,36 @@ void IApplicationFunctions::GetSaveDataSize(HLERequestContext& ctx) {
     rb.Push(size.journal);
 }
 
+void IApplicationFunctions::CreateCacheStorage(HLERequestContext& ctx) {
+    struct InputParameters {
+        u16 index;
+        s64 size;
+        s64 journal_size;
+    };
+    static_assert(sizeof(InputParameters) == 24);
+
+    struct OutputParameters {
+        u32 storage_target;
+        u64 required_size;
+    };
+    static_assert(sizeof(OutputParameters) == 16);
+
+    IPC::RequestParser rp{ctx};
+    const auto params = rp.PopRaw<InputParameters>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called with index={}, size={:#x}, journal_size={:#x}",
+                params.index, params.size, params.journal_size);
+
+    const OutputParameters resp{
+        .storage_target = 1,
+        .required_size = 0,
+    };
+
+    IPC::ResponseBuilder rb{ctx, 6};
+    rb.Push(ResultSuccess);
+    rb.PushRaw(resp);
+}
+
 void IApplicationFunctions::QueryApplicationPlayStatistics(HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
 

src/core/hle/service/am/am.h

@@ -333,6 +333,7 @@ private:
     void GetPseudoDeviceId(HLERequestContext& ctx);
     void ExtendSaveData(HLERequestContext& ctx);
     void GetSaveDataSize(HLERequestContext& ctx);
+    void CreateCacheStorage(HLERequestContext& ctx);
     void BeginBlockingHomeButtonShortAndLongPressed(HLERequestContext& ctx);
     void EndBlockingHomeButtonShortAndLongPressed(HLERequestContext& ctx);
     void BeginBlockingHomeButton(HLERequestContext& ctx);

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -24,8 +24,10 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/system_archive/system_archive.h"
 #include "core/file_sys/vfs.h"
+#include "core/hle/result.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/filesystem/fsp_srv.h"
+#include "core/hle/service/hle_ipc.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/reporter.h"
 
@@ -552,9 +554,9 @@ public:
         // Write the data to memory
         ctx.WriteBuffer(begin, range_size);
 
-        IPC::ResponseBuilder rb{ctx, 3};
+        IPC::ResponseBuilder rb{ctx, 4};
         rb.Push(ResultSuccess);
-        rb.Push<u32>(static_cast<u32>(actual_entries));
+        rb.Push<u64>(actual_entries);
     }
 
 private:
@@ -712,7 +714,7 @@ FSP_SRV::FSP_SRV(Core::System& system_)
         {59, nullptr, "WriteSaveDataFileSystemExtraData"},
         {60, nullptr, "OpenSaveDataInfoReader"},
         {61, &FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId, "OpenSaveDataInfoReaderBySaveDataSpaceId"},
-        {62, nullptr, "OpenCacheStorageList"},
+        {62, &FSP_SRV::OpenSaveDataInfoReaderOnlyCacheStorage, "OpenSaveDataInfoReaderOnlyCacheStorage"},
         {64, nullptr, "OpenSaveDataInternalStorageFileSystem"},
         {65, nullptr, "UpdateSaveDataMacForDebug"},
         {66, nullptr, "WriteSaveDataFileSystemExtraData2"},
@@ -921,6 +923,15 @@ void FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId(HLERequestContext& ctx) {
         std::make_shared<ISaveDataInfoReader>(system, space, fsc));
 }
 
+void FSP_SRV::OpenSaveDataInfoReaderOnlyCacheStorage(HLERequestContext& ctx) {
+    LOG_WARNING(Service_FS, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(ResultSuccess);
+    rb.PushIpcInterface<ISaveDataInfoReader>(system, FileSys::SaveDataSpaceId::TemporaryStorage,
+                                             fsc);
+}
+
 void FSP_SRV::WriteSaveDataFileSystemExtraDataBySaveDataAttribute(HLERequestContext& ctx) {
     LOG_WARNING(Service_FS, "(STUBBED) called.");
 

src/core/hle/service/filesystem/fsp_srv.h

@@ -42,6 +42,7 @@ private:
     void OpenSaveDataFileSystem(HLERequestContext& ctx);
     void OpenReadOnlySaveDataFileSystem(HLERequestContext& ctx);
     void OpenSaveDataInfoReaderBySaveDataSpaceId(HLERequestContext& ctx);
+    void OpenSaveDataInfoReaderOnlyCacheStorage(HLERequestContext& ctx);
     void WriteSaveDataFileSystemExtraDataBySaveDataAttribute(HLERequestContext& ctx);
     void ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute(HLERequestContext& ctx);
     void OpenDataStorageByCurrentProcess(HLERequestContext& ctx);

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

@@ -423,8 +423,8 @@ void Controller_NPad::RequestPadStateUpdate(Core::HID::NpadIdType npad_id) {
         return;
     }
 
-    // This function is unique to yuzu for the turbo buttons to work properly
-    controller.device->TurboButtonUpdate();
+    // This function is unique to yuzu for the turbo buttons and motion to work properly
+    controller.device->StatusUpdate();
 
     auto& pad_entry = controller.npad_pad_state;
     auto& trigger_entry = controller.npad_trigger_state;
@@ -979,8 +979,8 @@ void Controller_NPad::VibrateController(
 }
 
 void Controller_NPad::VibrateControllers(
-    const std::vector<Core::HID::VibrationDeviceHandle>& vibration_device_handles,
-    const std::vector<Core::HID::VibrationValue>& vibration_values) {
+    std::span<const Core::HID::VibrationDeviceHandle> vibration_device_handles,
+    std::span<const Core::HID::VibrationValue> vibration_values) {
     if (!Settings::values.vibration_enabled.GetValue() && !permit_vibration_session_enabled) {
         return;
     }

src/core/hle/service/hid/controllers/npad.h

@@ -112,8 +112,8 @@ public:
                            const Core::HID::VibrationValue& vibration_value);
 
     void VibrateControllers(
-        const std::vector<Core::HID::VibrationDeviceHandle>& vibration_device_handles,
-        const std::vector<Core::HID::VibrationValue>& vibration_values);
+        std::span<const Core::HID::VibrationDeviceHandle> vibration_device_handles,
+        std::span<const Core::HID::VibrationValue> vibration_values);
 
     Core::HID::VibrationValue GetLastVibration(
         const Core::HID::VibrationDeviceHandle& vibration_device_handle) const;

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

@@ -1601,16 +1601,16 @@ void Hid::SendVibrationValues(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto applet_resource_user_id{rp.Pop<u64>()};
 
-    const auto handles = ctx.ReadBuffer(0);
-    const auto vibrations = ctx.ReadBuffer(1);
+    const auto handle_data = ctx.ReadBuffer(0);
+    const auto handle_count = ctx.GetReadBufferNumElements<Core::HID::VibrationDeviceHandle>(0);
+    const auto vibration_data = ctx.ReadBuffer(1);
+    const auto vibration_count = ctx.GetReadBufferNumElements<Core::HID::VibrationValue>(1);
 
-    std::vector<Core::HID::VibrationDeviceHandle> vibration_device_handles(
-        handles.size() / sizeof(Core::HID::VibrationDeviceHandle));
-    std::vector<Core::HID::VibrationValue> vibration_values(vibrations.size() /
-                                                            sizeof(Core::HID::VibrationValue));
-
-    std::memcpy(vibration_device_handles.data(), handles.data(), handles.size());
-    std::memcpy(vibration_values.data(), vibrations.data(), vibrations.size());
+    auto vibration_device_handles =
+        std::span(reinterpret_cast<const Core::HID::VibrationDeviceHandle*>(handle_data.data()),
+                  handle_count);
+    auto vibration_values = std::span(
+        reinterpret_cast<const Core::HID::VibrationValue*>(vibration_data.data()), vibration_count);
 
     applet_resource->GetController<Controller_NPad>(HidController::NPad)
         .VibrateControllers(vibration_device_handles, vibration_values);

src/core/hle/service/nfc/common/amiibo_crypto.cpp

@@ -52,9 +52,6 @@ bool IsAmiiboValid(const EncryptedNTAG215File& ntag_file) {
     if (ntag_file.compability_container != 0xEEFF10F1U) {
         return false;
     }
-    if (amiibo_data.constant_value != 0xA5) {
-        return false;
-    }
     if (amiibo_data.model_info.tag_type != NFC::PackedTagType::Type2) {
         return false;
     }

src/core/hle/service/nfc/common/device.cpp

@@ -48,9 +48,6 @@ NfcDevice::NfcDevice(Core::HID::NpadIdType npad_id_, Core::System& system_,
     };
     is_controller_set = true;
     callback_key = npad_device->SetCallback(engine_callback);
-
-    auto& standard_steady_clock{system.GetTimeManager().GetStandardSteadyClockCore()};
-    current_posix_time = standard_steady_clock.GetCurrentTimePoint(system).time_point;
 }
 
 NfcDevice::~NfcDevice() {
@@ -122,18 +119,31 @@ bool NfcDevice::LoadNfcTag(std::span<const u8> data) {
 
     memcpy(&tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
     is_plain_amiibo = NFP::AmiiboCrypto::IsAmiiboValid(tag_data);
-
-    if (is_plain_amiibo) {
-        encrypted_tag_data = NFP::AmiiboCrypto::EncodedDataToNfcData(tag_data);
-        LOG_INFO(Service_NFP, "Using plain amiibo");
-    } else {
-        tag_data = {};
-        memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
-    }
+    is_write_protected = false;
 
     device_state = DeviceState::TagFound;
     deactivate_event->GetReadableEvent().Clear();
     activate_event->Signal();
+
+    // Fallback for plain amiibos
+    if (is_plain_amiibo) {
+        LOG_INFO(Service_NFP, "Using plain amiibo");
+        encrypted_tag_data = NFP::AmiiboCrypto::EncodedDataToNfcData(tag_data);
+        return true;
+    }
+
+    // Fallback for encrypted amiibos without keys
+    if (!NFP::AmiiboCrypto::IsKeyAvailable()) {
+        LOG_INFO(Service_NFC, "Loading amiibo without keys");
+        memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
+        BuildAmiiboWithoutKeys();
+        is_plain_amiibo = true;
+        is_write_protected = true;
+        return true;
+    }
+
+    tag_data = {};
+    memcpy(&encrypted_tag_data, data.data(), sizeof(NFP::EncryptedNTAG215File));
     return true;
 }
 
@@ -227,11 +237,21 @@ Result NfcDevice::GetTagInfo(NFP::TagInfo& tag_info, bool is_mifare) const {
         return ResultWrongDeviceState;
     }
 
+    UniqueSerialNumber uuid = encrypted_tag_data.uuid.uid;
+
+    // Generate random UUID to bypass amiibo load limits
+    if (Settings::values.random_amiibo_id) {
+        Common::TinyMT rng{};
+        rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
+        rng.GenerateRandomBytes(uuid.data(), sizeof(UniqueSerialNumber));
+        uuid[3] = 0x88 ^ uuid[0] ^ uuid[1] ^ uuid[2];
+    }
+
     if (is_mifare) {
         tag_info = {
-            .uuid = encrypted_tag_data.uuid.uid,
+            .uuid = uuid,
             .uuid_extension = {},
-            .uuid_length = static_cast<u8>(encrypted_tag_data.uuid.uid.size()),
+            .uuid_length = static_cast<u8>(uuid.size()),
             .protocol = NfcProtocol::TypeA,
             .tag_type = TagType::Type4,
         };
@@ -240,9 +260,9 @@ Result NfcDevice::GetTagInfo(NFP::TagInfo& tag_info, bool is_mifare) const {
 
     // Protocol and tag type may change here
     tag_info = {
-        .uuid = encrypted_tag_data.uuid.uid,
+        .uuid = uuid,
         .uuid_extension = {},
-        .uuid_length = static_cast<u8>(encrypted_tag_data.uuid.uid.size()),
+        .uuid_length = static_cast<u8>(uuid.size()),
         .protocol = NfcProtocol::TypeA,
         .tag_type = TagType::Type2,
     };
@@ -339,23 +359,15 @@ Result NfcDevice::Mount(NFP::ModelType model_type, NFP::MountTarget mount_target
         return ResultWrongDeviceState;
     }
 
-    // The loaded amiibo is not encrypted
-    if (is_plain_amiibo) {
-        device_state = DeviceState::TagMounted;
-        mount_target = mount_target_;
-        return ResultSuccess;
-    }
-
     if (!NFP::AmiiboCrypto::IsAmiiboValid(encrypted_tag_data)) {
         LOG_ERROR(Service_NFP, "Not an amiibo");
         return ResultNotAnAmiibo;
     }
 
-    // Mark amiibos as read only when keys are missing
-    if (!NFP::AmiiboCrypto::IsKeyAvailable()) {
-        LOG_ERROR(Service_NFP, "No keys detected");
+    // The loaded amiibo is not encrypted
+    if (is_plain_amiibo) {
         device_state = DeviceState::TagMounted;
-        mount_target = NFP::MountTarget::Rom;
+        mount_target = mount_target_;
         return ResultSuccess;
     }
 
@@ -406,7 +418,7 @@ Result NfcDevice::Flush() {
 
     auto& settings = tag_data.settings;
 
-    const auto& current_date = GetAmiiboDate(current_posix_time);
+    const auto& current_date = GetAmiiboDate(GetCurrentPosixTime());
     if (settings.write_date.raw_date != current_date.raw_date) {
         settings.write_date = current_date;
         UpdateSettingsCrc();
@@ -414,11 +426,11 @@ Result NfcDevice::Flush() {
 
     tag_data.write_counter++;
 
-    FlushWithBreak(NFP::BreakType::Normal);
+    const auto result = FlushWithBreak(NFP::BreakType::Normal);
 
     is_data_moddified = false;
 
-    return ResultSuccess;
+    return result;
 }
 
 Result NfcDevice::FlushDebug() {
@@ -437,11 +449,11 @@ Result NfcDevice::FlushDebug() {
 
     tag_data.write_counter++;
 
-    FlushWithBreak(NFP::BreakType::Normal);
+    const auto result = FlushWithBreak(NFP::BreakType::Normal);
 
     is_data_moddified = false;
 
-    return ResultSuccess;
+    return result;
 }
 
 Result NfcDevice::FlushWithBreak(NFP::BreakType break_type) {
@@ -450,6 +462,11 @@ Result NfcDevice::FlushWithBreak(NFP::BreakType break_type) {
         return ResultWrongDeviceState;
     }
 
+    if (is_write_protected) {
+        LOG_ERROR(Service_NFP, "No keys available skipping write request");
+        return ResultSuccess;
+    }
+
     std::vector<u8> data(sizeof(NFP::EncryptedNTAG215File));
     if (is_plain_amiibo) {
         memcpy(data.data(), &tag_data, sizeof(tag_data));
@@ -525,6 +542,7 @@ Result NfcDevice::GetModelInfo(NFP::ModelInfo& model_info) const {
     }
 
     const auto& model_info_data = encrypted_tag_data.user_memory.model_info;
+
     model_info = {
         .character_id = model_info_data.character_id,
         .character_variant = model_info_data.character_variant,
@@ -669,6 +687,7 @@ Result NfcDevice::DeleteRegisterInfo() {
     }
 
     Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
     rng.GenerateRandomBytes(&tag_data.owner_mii, sizeof(tag_data.owner_mii));
     rng.GenerateRandomBytes(&tag_data.settings.amiibo_name, sizeof(tag_data.settings.amiibo_name));
     rng.GenerateRandomBytes(&tag_data.unknown, sizeof(u8));
@@ -701,7 +720,7 @@ Result NfcDevice::SetRegisterInfoPrivate(const NFP::RegisterInfoPrivate& registe
     auto& settings = tag_data.settings;
 
     if (tag_data.settings.settings.amiibo_initialized == 0) {
-        settings.init_date = GetAmiiboDate(current_posix_time);
+        settings.init_date = GetAmiiboDate(GetCurrentPosixTime());
         settings.write_date.raw_date = 0;
     }
 
@@ -868,6 +887,7 @@ Result NfcDevice::SetApplicationArea(std::span<const u8> data) {
     }
 
     Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
     std::memcpy(tag_data.application_area.data(), data.data(), data.size());
     // Fill remaining data with random numbers
     rng.GenerateRandomBytes(tag_data.application_area.data() + data.size(),
@@ -924,6 +944,7 @@ Result NfcDevice::RecreateApplicationArea(u32 access_id, std::span<const u8> dat
     }
 
     Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
     std::memcpy(tag_data.application_area.data(), data.data(), data.size());
     // Fill remaining data with random numbers
     rng.GenerateRandomBytes(tag_data.application_area.data() + data.size(),
@@ -973,6 +994,7 @@ Result NfcDevice::DeleteApplicationArea() {
     }
 
     Common::TinyMT rng{};
+    rng.Initialize(static_cast<u32>(GetCurrentPosixTime()));
     rng.GenerateRandomBytes(tag_data.application_area.data(), sizeof(NFP::ApplicationArea));
     rng.GenerateRandomBytes(&tag_data.application_id, sizeof(u64));
     rng.GenerateRandomBytes(&tag_data.application_area_id, sizeof(u32));
@@ -1021,7 +1043,6 @@ Result NfcDevice::GetAll(NFP::NfpData& data) const {
     }
 
     NFP::CommonInfo common_info{};
-    Service::Mii::MiiManager manager;
     const u64 application_id = tag_data.application_id;
 
     GetCommonInfo(common_info);
@@ -1189,6 +1210,11 @@ NFP::AmiiboDate NfcDevice::GetAmiiboDate(s64 posix_time) const {
     return amiibo_date;
 }
 
+u64 NfcDevice::GetCurrentPosixTime() const {
+    auto& standard_steady_clock{system.GetTimeManager().GetStandardSteadyClockCore()};
+    return standard_steady_clock.GetCurrentTimePoint(system).time_point;
+}
+
 u64 NfcDevice::RemoveVersionByte(u64 application_id) const {
     return application_id & ~(0xfULL << NFP::application_id_version_offset);
 }
@@ -1232,6 +1258,28 @@ void NfcDevice::UpdateRegisterInfoCrc() {
     tag_data.register_info_crc = crc.checksum();
 }
 
+void NfcDevice::BuildAmiiboWithoutKeys() {
+    Service::Mii::MiiManager manager;
+    auto& settings = tag_data.settings;
+
+    tag_data = NFP::AmiiboCrypto::NfcDataToEncodedData(encrypted_tag_data);
+
+    // Common info
+    tag_data.write_counter = 0;
+    tag_data.amiibo_version = 0;
+    settings.write_date = GetAmiiboDate(GetCurrentPosixTime());
+
+    // Register info
+    SetAmiiboName(settings, {'y', 'u', 'z', 'u', 'A', 'm', 'i', 'i', 'b', 'o'});
+    settings.settings.font_region.Assign(0);
+    settings.init_date = GetAmiiboDate(GetCurrentPosixTime());
+    tag_data.owner_mii = manager.BuildFromStoreData(manager.BuildDefault(0));
+
+    // Admin info
+    settings.settings.amiibo_initialized.Assign(1);
+    settings.settings.appdata_initialized.Assign(0);
+}
+
 u64 NfcDevice::GetHandle() const {
     // Generate a handle based of the npad id
     return static_cast<u64>(npad_id);

src/core/hle/service/nfc/common/device.h

@@ -105,10 +105,13 @@ private:
     NFP::AmiiboName GetAmiiboName(const NFP::AmiiboSettings& settings) const;
     void SetAmiiboName(NFP::AmiiboSettings& settings, const NFP::AmiiboName& amiibo_name);
     NFP::AmiiboDate GetAmiiboDate(s64 posix_time) const;
+    u64 GetCurrentPosixTime() const;
     u64 RemoveVersionByte(u64 application_id) const;
     void UpdateSettingsCrc();
     void UpdateRegisterInfoCrc();
 
+    void BuildAmiiboWithoutKeys();
+
     bool is_controller_set{};
     int callback_key;
     const Core::HID::NpadIdType npad_id;
@@ -127,7 +130,7 @@ private:
     bool is_data_moddified{};
     bool is_app_area_open{};
     bool is_plain_amiibo{};
-    s64 current_posix_time{};
+    bool is_write_protected{};
     NFP::MountTarget mount_target{NFP::MountTarget::None};
 
     NFP::NTAG215File tag_data{};

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

@@ -218,7 +218,7 @@ public:
 
 private:
     void Submit(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         if (state == RequestState::NotSubmitted) {
             UpdateState(RequestState::OnHold);
@@ -229,7 +229,7 @@ private:
     }
 
     void GetRequestState(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(ResultSuccess);
@@ -237,7 +237,7 @@ private:
     }
 
     void GetResult(HLERequestContext& ctx) {
-        LOG_WARNING(Service_NIFM, "(STUBBED) called");
+        LOG_DEBUG(Service_NIFM, "(STUBBED) called");
 
         const auto result = [this] {
             const auto has_connection = Network::GetHostIPv4Address().has_value();

src/core/hle/service/nvnflinger/buffer_queue_producer.cpp

@@ -793,6 +793,7 @@ Status BufferQueueProducer::SetPreallocatedBuffer(s32 slot,
     std::scoped_lock lock{core->mutex};
 
     slots[slot] = {};
+    slots[slot].fence = Fence::NoFence();
     slots[slot].graphic_buffer = buffer;
     slots[slot].frame_number = 0;
 
@@ -854,7 +855,7 @@ void BufferQueueProducer::Transact(HLERequestContext& ctx, TransactionId code, u
         status = DequeueBuffer(&slot, &fence, is_async, width, height, pixel_format, usage);
 
         parcel_out.Write(slot);
-        parcel_out.WriteObject(&fence);
+        parcel_out.WriteFlattenedObject(&fence);
         break;
     }
     case TransactionId::RequestBuffer: {
@@ -864,7 +865,7 @@ void BufferQueueProducer::Transact(HLERequestContext& ctx, TransactionId code, u
 
         status = RequestBuffer(slot, &buf);
 
-        parcel_out.WriteObject(buf);
+        parcel_out.WriteFlattenedObject(buf);
         break;
     }
     case TransactionId::QueueBuffer: {

src/core/hle/service/nvnflinger/parcel.h

@@ -117,61 +117,67 @@ private:
 
 class OutputParcel final {
 public:
-    static constexpr std::size_t DefaultBufferSize = 0x40;
-
-    OutputParcel() : buffer(DefaultBufferSize) {}
-
-    template <typename T>
-    explicit OutputParcel(const T& out_data) : buffer(DefaultBufferSize) {
-        Write(out_data);
-    }
+    OutputParcel() = default;
 
     template <typename T>
     void Write(const T& val) {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable.");
-
-        if (buffer.size() < write_index + sizeof(T)) {
-            buffer.resize(buffer.size() + sizeof(T) + DefaultBufferSize);
-        }
-
-        std::memcpy(buffer.data() + write_index, &val, sizeof(T));
-        write_index += sizeof(T);
-        write_index = Common::AlignUp(write_index, 4);
+        this->WriteImpl(val, m_data_buffer);
     }
 
     template <typename T>
-    void WriteObject(const T* ptr) {
-        static_assert(std::is_trivially_copyable_v<T>, "T must be trivially copyable.");
-
+    void WriteFlattenedObject(const T* ptr) {
         if (!ptr) {
-            Write<u32>(0);
+            this->Write<u32>(0);
             return;
         }
 
-        Write<u32>(1);
-        Write<s64>(sizeof(T));
-        Write(*ptr);
+        this->Write<u32>(1);
+        this->Write<s64>(sizeof(T));
+        this->Write(*ptr);
     }
 
     template <typename T>
-    void WriteObject(const std::shared_ptr<T> ptr) {
-        WriteObject(ptr.get());
+    void WriteFlattenedObject(const std::shared_ptr<T> ptr) {
+        this->WriteFlattenedObject(ptr.get());
+    }
+
+    template <typename T>
+    void WriteInterface(const T& val) {
+        this->WriteImpl(val, m_data_buffer);
+        this->WriteImpl(0U, m_object_buffer);
     }
 
     std::vector<u8> Serialize() const {
-        ParcelHeader header{};
-        header.data_size = static_cast<u32>(write_index - sizeof(ParcelHeader));
-        header.data_offset = sizeof(ParcelHeader);
-        header.objects_size = 4;
-        header.objects_offset = static_cast<u32>(sizeof(ParcelHeader) + header.data_size);
-        std::memcpy(buffer.data(), &header, sizeof(ParcelHeader));
+        std::vector<u8> output_buffer(sizeof(ParcelHeader) + m_data_buffer.size() +
+                                      m_object_buffer.size());
 
-        return buffer;
+        ParcelHeader header{};
+        header.data_size = static_cast<u32>(m_data_buffer.size());
+        header.data_offset = sizeof(ParcelHeader);
+        header.objects_size = static_cast<u32>(m_object_buffer.size());
+        header.objects_offset = header.data_offset + header.data_size;
+
+        std::memcpy(output_buffer.data(), &header, sizeof(header));
+        std::ranges::copy(m_data_buffer, output_buffer.data() + header.data_offset);
+        std::ranges::copy(m_object_buffer, output_buffer.data() + header.objects_offset);
+
+        return output_buffer;
     }
 
 private:
-    mutable std::vector<u8> buffer;
-    std::size_t write_index = sizeof(ParcelHeader);
+    template <typename T>
+        requires(std::is_trivially_copyable_v<T>)
+    void WriteImpl(const T& val, std::vector<u8>& buffer) {
+        const size_t aligned_size = Common::AlignUp(sizeof(T), 4);
+        const size_t old_size = buffer.size();
+        buffer.resize(old_size + aligned_size);
+
+        std::memcpy(buffer.data() + old_size, &val, sizeof(T));
+    }
+
+private:
+    std::vector<u8> m_data_buffer;
+    std::vector<u8> m_object_buffer;
 };
 
 } // namespace Service::android

src/core/hle/service/time/clock_types.h

@@ -59,6 +59,18 @@ static_assert(sizeof(SystemClockContext) == 0x20, "SystemClockContext is incorre
 static_assert(std::is_trivially_copyable_v<SystemClockContext>,
               "SystemClockContext must be trivially copyable");
 
+struct ContinuousAdjustmentTimePoint {
+    s64 measurement_offset;
+    s64 diff_scale;
+    u32 shift_amount;
+    s64 lower;
+    s64 upper;
+    Common::UUID clock_source_id;
+};
+static_assert(sizeof(ContinuousAdjustmentTimePoint) == 0x38);
+static_assert(std::is_trivially_copyable_v<ContinuousAdjustmentTimePoint>,
+              "ContinuousAdjustmentTimePoint must be trivially copyable");
+
 /// https://switchbrew.org/wiki/Glue_services#TimeSpanType
 struct TimeSpanType {
     s64 nanoseconds{};

src/core/hle/service/time/time_sharedmemory.cpp

@@ -30,6 +30,25 @@ void SharedMemory::SetupStandardSteadyClock(const Common::UUID& clock_source_id,
 }
 
 void SharedMemory::UpdateLocalSystemClockContext(const Clock::SystemClockContext& context) {
+    // lower and upper are related to the measurement point for the steady time point,
+    // and compare equal on boot
+    const s64 time_point_ns = context.steady_time_point.time_point * 1'000'000'000LL;
+
+    // This adjusts for some sort of time skew
+    // Both 0 on boot
+    const s64 diff_scale = 0;
+    const u32 shift_amount = 0;
+
+    const Clock::ContinuousAdjustmentTimePoint adjustment{
+        .measurement_offset = system.CoreTiming().GetGlobalTimeNs().count(),
+        .diff_scale = diff_scale,
+        .shift_amount = shift_amount,
+        .lower = time_point_ns,
+        .upper = time_point_ns,
+        .clock_source_id = context.steady_time_point.clock_source_id,
+    };
+
+    StoreToLockFreeAtomicType(&GetFormat()->continuous_adjustment_timepoint, adjustment);
     StoreToLockFreeAtomicType(&GetFormat()->standard_local_system_clock_context, context);
 }
 

src/core/hle/service/time/time_sharedmemory.h

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

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

@@ -64,8 +64,8 @@ public:
 private:
     const u32 magic = 2;
     const u32 process_id = 1;
-    const u32 id;
-    INSERT_PADDING_WORDS(3);
+    const u64 id;
+    INSERT_PADDING_WORDS(2);
     std::array<u8, 8> dispdrv = {'d', 'i', 's', 'p', 'd', 'r', 'v', '\0'};
     INSERT_PADDING_WORDS(2);
 };
@@ -608,7 +608,9 @@ private:
             return;
         }
 
-        const auto parcel = android::OutputParcel{NativeWindow{*buffer_queue_id}};
+        android::OutputParcel parcel;
+        parcel.WriteInterface(NativeWindow{*buffer_queue_id});
+
         const auto buffer_size = ctx.WriteBuffer(parcel.Serialize());
 
         IPC::ResponseBuilder rb{ctx, 4};
@@ -654,7 +656,9 @@ private:
             return;
         }
 
-        const auto parcel = android::OutputParcel{NativeWindow{*buffer_queue_id}};
+        android::OutputParcel parcel;
+        parcel.WriteInterface(NativeWindow{*buffer_queue_id});
+
         const auto buffer_size = ctx.WriteBuffer(parcel.Serialize());
 
         IPC::ResponseBuilder rb{ctx, 6};

src/core/internal_network/network.cpp

@@ -356,7 +356,7 @@ NetworkInstance::~NetworkInstance() {
 std::optional<IPv4Address> GetHostIPv4Address() {
     const auto network_interface = Network::GetSelectedNetworkInterface();
     if (!network_interface.has_value()) {
-        LOG_ERROR(Network, "GetSelectedNetworkInterface returned no interface");
+        LOG_DEBUG(Network, "GetSelectedNetworkInterface returned no interface");
         return {};
     }
 

src/core/internal_network/network_interface.cpp

@@ -200,7 +200,7 @@ std::optional<NetworkInterface> GetSelectedNetworkInterface() {
         });
 
     if (res == network_interfaces.end()) {
-        LOG_ERROR(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
+        LOG_DEBUG(Network, "Couldn't find selected interface \"{}\"", selected_network_interface);
         return std::nullopt;
     }
 

src/core/memory.cpp

@@ -13,10 +13,12 @@
 #include "common/swap.h"
 #include "core/core.h"
 #include "core/device_memory.h"
+#include "core/hardware_properties.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Core::Memory {
 
@@ -243,7 +245,7 @@ struct Memory::Impl {
             [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount,
                 const u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
-                    system.GPU().FlushRegion(GetInteger(current_vaddr), copy_amount);
+                    HandleRasterizerDownload(GetInteger(current_vaddr), copy_amount);
                 }
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
             },
@@ -334,7 +336,7 @@ struct Memory::Impl {
             },
             [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount,
                 u8* const host_ptr) {
-                system.GPU().FlushRegion(GetInteger(current_vaddr), copy_amount);
+                HandleRasterizerDownload(GetInteger(current_vaddr), copy_amount);
                 WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount);
             },
             [&](const std::size_t copy_amount) {
@@ -373,7 +375,7 @@ struct Memory::Impl {
                                  const std::size_t block_size) {
             // dc ivac: Invalidate to point of coherency
             // GPU flush -> CPU invalidate
-            system.GPU().FlushRegion(GetInteger(current_vaddr), block_size);
+            HandleRasterizerDownload(GetInteger(current_vaddr), block_size);
         };
         return PerformCacheOperation(process, dest_addr, size, on_rasterizer);
     }
@@ -462,7 +464,8 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
+            const bool is_read_enable =
+                !Settings::values.use_reactive_flushing.GetValue() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 
@@ -651,7 +654,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8,
                           GetInteger(vaddr));
             },
-            [&]() { system.GPU().FlushRegion(GetInteger(vaddr), sizeof(T)); });
+            [&]() { HandleRasterizerDownload(GetInteger(vaddr), sizeof(T)); });
         if (ptr) {
             std::memcpy(&result, ptr, sizeof(T));
         }
@@ -712,7 +715,19 @@ struct Memory::Impl {
         return true;
     }
 
+    void HandleRasterizerDownload(VAddr address, size_t size) {
+        const size_t core = system.GetCurrentHostThreadID();
+        auto& current_area = rasterizer_areas[core];
+        const VAddr end_address = address + size;
+        if (current_area.start_address <= address && end_address <= current_area.end_address)
+            [[likely]] {
+            return;
+        }
+        current_area = system.GPU().OnCPURead(address, size);
+    }
+
     Common::PageTable* current_page_table = nullptr;
+    std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> rasterizer_areas{};
     Core::System& system;
 };
 

src/input_common/drivers/joycon.cpp

@@ -195,8 +195,8 @@ void Joycons::RegisterNewDevice(SDL_hid_device_info* device_info) {
                 OnMotionUpdate(port, type, id, value);
             }},
             .on_ring_data = {[this](f32 ring_data) { OnRingConUpdate(ring_data); }},
-            .on_amiibo_data = {[this, port](const std::vector<u8>& amiibo_data) {
-                OnAmiiboUpdate(port, amiibo_data);
+            .on_amiibo_data = {[this, port, type](const std::vector<u8>& amiibo_data) {
+                OnAmiiboUpdate(port, type, amiibo_data);
             }},
             .on_camera_data = {[this, port](const std::vector<u8>& camera_data,
                                             Joycon::IrsResolution format) {
@@ -291,9 +291,13 @@ Common::Input::NfcState Joycons::SupportsNfc(const PadIdentifier& identifier_) c
     return Common::Input::NfcState::Success;
 };
 
-Common::Input::NfcState Joycons::WriteNfcData(const PadIdentifier& identifier_,
+Common::Input::NfcState Joycons::WriteNfcData(const PadIdentifier& identifier,
                                               const std::vector<u8>& data) {
-    return Common::Input::NfcState::NotSupported;
+    auto handle = GetHandle(identifier);
+    if (handle->WriteNfcData(data) != Joycon::DriverResult::Success) {
+        return Common::Input::NfcState::WriteFailed;
+    }
+    return Common::Input::NfcState::Success;
 };
 
 Common::Input::DriverResult Joycons::SetPollingMode(const PadIdentifier& identifier,
@@ -398,8 +402,9 @@ void Joycons::OnRingConUpdate(f32 ring_data) {
     SetAxis(identifier, 100, ring_data);
 }
 
-void Joycons::OnAmiiboUpdate(std::size_t port, const std::vector<u8>& amiibo_data) {
-    const auto identifier = GetIdentifier(port, Joycon::ControllerType::Right);
+void Joycons::OnAmiiboUpdate(std::size_t port, Joycon::ControllerType type,
+                             const std::vector<u8>& amiibo_data) {
+    const auto identifier = GetIdentifier(port, type);
     const auto nfc_state = amiibo_data.empty() ? Common::Input::NfcState::AmiiboRemoved
                                                : Common::Input::NfcState::NewAmiibo;
     SetNfc(identifier, {nfc_state, amiibo_data});

src/input_common/drivers/joycon.h

@@ -81,7 +81,8 @@ private:
     void OnMotionUpdate(std::size_t port, Joycon::ControllerType type, int id,
                         const Joycon::MotionData& value);
     void OnRingConUpdate(f32 ring_data);
-    void OnAmiiboUpdate(std::size_t port, const std::vector<u8>& amiibo_data);
+    void OnAmiiboUpdate(std::size_t port, Joycon::ControllerType type,
+                        const std::vector<u8>& amiibo_data);
     void OnCameraUpdate(std::size_t port, const std::vector<u8>& camera_data,
                         Joycon::IrsResolution format);
 

src/input_common/drivers/sdl_driver.cpp

@@ -109,14 +109,37 @@ public:
     }
 
     bool RumblePlay(const Common::Input::VibrationStatus vibration) {
-        constexpr u32 rumble_max_duration_ms = 1000;
+        constexpr u32 rumble_max_duration_ms = 2000;
+        constexpr f32 low_start_sensitivity_limit = 140.0;
+        constexpr f32 low_width_sensitivity_limit = 400.0;
+        constexpr f32 high_start_sensitivity_limit = 200.0;
+        constexpr f32 high_width_sensitivity_limit = 700.0;
+        // Try to provide some feeling of the frequency by reducing the amplitude depending on it.
+        f32 low_frequency_scale = 1.0;
+        if (vibration.low_frequency > low_start_sensitivity_limit) {
+            low_frequency_scale =
+                std::max(1.0f - (vibration.low_frequency - low_start_sensitivity_limit) /
+                                    low_width_sensitivity_limit,
+                         0.3f);
+        }
+        f32 low_amplitude = vibration.low_amplitude * low_frequency_scale;
+
+        f32 high_frequency_scale = 1.0;
+        if (vibration.high_frequency > high_start_sensitivity_limit) {
+            high_frequency_scale =
+                std::max(1.0f - (vibration.high_frequency - high_start_sensitivity_limit) /
+                                    high_width_sensitivity_limit,
+                         0.3f);
+        }
+        f32 high_amplitude = vibration.high_amplitude * high_frequency_scale;
+
         if (sdl_controller) {
-            return SDL_GameControllerRumble(
-                       sdl_controller.get(), static_cast<u16>(vibration.low_amplitude),
-                       static_cast<u16>(vibration.high_amplitude), rumble_max_duration_ms) != -1;
+            return SDL_GameControllerRumble(sdl_controller.get(), static_cast<u16>(low_amplitude),
+                                            static_cast<u16>(high_amplitude),
+                                            rumble_max_duration_ms) != -1;
         } else if (sdl_joystick) {
-            return SDL_JoystickRumble(sdl_joystick.get(), static_cast<u16>(vibration.low_amplitude),
-                                      static_cast<u16>(vibration.high_amplitude),
+            return SDL_JoystickRumble(sdl_joystick.get(), static_cast<u16>(low_amplitude),
+                                      static_cast<u16>(high_amplitude),
                                       rumble_max_duration_ms) != -1;
         }
 

src/input_common/helpers/joycon_driver.cpp

@@ -492,6 +492,26 @@ DriverResult JoyconDriver::SetRingConMode() {
     return result;
 }
 
+DriverResult JoyconDriver::WriteNfcData(std::span<const u8> data) {
+    std::scoped_lock lock{mutex};
+    disable_input_thread = true;
+
+    if (!supported_features.nfc) {
+        return DriverResult::NotSupported;
+    }
+    if (!nfc_protocol->IsEnabled()) {
+        return DriverResult::Disabled;
+    }
+    if (!amiibo_detected) {
+        return DriverResult::ErrorWritingData;
+    }
+
+    const auto result = nfc_protocol->WriteAmiibo(data);
+
+    disable_input_thread = false;
+    return result;
+}
+
 bool JoyconDriver::IsConnected() const {
     std::scoped_lock lock{mutex};
     return is_connected.load();

src/input_common/helpers/joycon_driver.h

@@ -49,6 +49,7 @@ public:
     DriverResult SetIrMode();
     DriverResult SetNfcMode();
     DriverResult SetRingConMode();
+    DriverResult WriteNfcData(std::span<const u8> data);
 
     void SetCallbacks(const JoyconCallbacks& callbacks);
 

src/input_common/helpers/joycon_protocol/common_protocol.cpp

@@ -236,13 +236,13 @@ DriverResult JoyconCommonProtocol::GetMCUDataResponse(ReportMode report_mode,
     return DriverResult::Success;
 }
 
-DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, SubCommand sc,
+DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, MCUSubCommand sc,
                                                std::span<const u8> buffer,
                                                MCUCommandResponse& output) {
     SubCommandPacket packet{
         .output_report = OutputReport::MCU_DATA,
         .packet_counter = GetCounter(),
-        .sub_command = sc,
+        .mcu_sub_command = sc,
         .command_data = {},
     };
 
@@ -265,12 +265,11 @@ DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, SubComman
 
 DriverResult JoyconCommonProtocol::WaitSetMCUMode(ReportMode report_mode, MCUMode mode) {
     MCUCommandResponse output{};
-    constexpr std::size_t MaxTries{8};
+    constexpr std::size_t MaxTries{16};
     std::size_t tries{};
 
     do {
-        const std::vector<u8> mcu_data{static_cast<u8>(MCUMode::Standby)};
-        const auto result = SendMCUData(report_mode, SubCommand::STATE, mcu_data, output);
+        const auto result = SendMCUData(report_mode, MCUSubCommand::SetDeviceMode, {}, output);
 
         if (result != DriverResult::Success) {
             return result;

src/input_common/helpers/joycon_protocol/common_protocol.h

@@ -156,7 +156,7 @@ public:
      * @param buffer data to be send
      * @returns output buffer containing the response
      */
-    DriverResult SendMCUData(ReportMode report_mode, SubCommand sc, std::span<const u8> buffer,
+    DriverResult SendMCUData(ReportMode report_mode, MCUSubCommand sc, std::span<const u8> buffer,
                              MCUCommandResponse& output);
 
     /**

src/input_common/helpers/joycon_protocol/joycon_types.h

@@ -23,6 +23,7 @@ constexpr std::array<u8, 8> DefaultVibrationBuffer{0x0, 0x1, 0x40, 0x40, 0x0, 0x
 
 using MacAddress = std::array<u8, 6>;
 using SerialNumber = std::array<u8, 15>;
+using TagUUID = std::array<u8, 7>;
 
 enum class ControllerType : u8 {
     None = 0x00,
@@ -276,12 +277,13 @@ enum class MCUPacketFlag : u8 {
     LastCommandPacket = 0x08,
 };
 
-enum class NFCReadCommand : u8 {
+enum class NFCCommand : u8 {
     CancelAll = 0x00,
     StartPolling = 0x01,
     StopPolling = 0x02,
     StartWaitingRecieve = 0x04,
-    Ntag = 0x06,
+    ReadNtag = 0x06,
+    WriteNtag = 0x08,
     Mifare = 0x0F,
 };
 
@@ -292,14 +294,19 @@ enum class NFCTagType : u8 {
 
 enum class NFCPages {
     Block0 = 0,
+    Block3 = 3,
     Block45 = 45,
     Block135 = 135,
     Block231 = 231,
 };
 
 enum class NFCStatus : u8 {
+    Ready = 0x00,
+    Polling = 0x01,
     LastPackage = 0x04,
+    WriteDone = 0x05,
     TagLost = 0x07,
+    WriteReady = 0x09,
 };
 
 enum class IrsMode : u8 {
@@ -394,6 +401,7 @@ enum class DriverResult {
     InvalidHandle,
     NotSupported,
     Disabled,
+    Delayed,
     Unknown,
 };
 
@@ -558,13 +566,32 @@ static_assert(sizeof(NFCReadBlockCommand) == 0x9, "NFCReadBlockCommand is an inv
 struct NFCReadCommandData {
     u8 unknown;
     u8 uuid_length;
-    u8 unknown_2;
-    std::array<u8, 6> uid;
+    TagUUID uid;
     NFCTagType tag_type;
     NFCReadBlockCommand read_block;
 };
 static_assert(sizeof(NFCReadCommandData) == 0x13, "NFCReadCommandData is an invalid size");
 
+#pragma pack(push, 1)
+struct NFCWriteCommandData {
+    u8 unknown;
+    u8 uuid_length;
+    TagUUID uid;
+    NFCTagType tag_type;
+    u8 unknown2;
+    u8 unknown3;
+    u8 unknown4;
+    u8 unknown5;
+    u8 unknown6;
+    u8 unknown7;
+    u8 unknown8;
+    u8 magic;
+    u16_be write_count;
+    u8 amiibo_version;
+};
+static_assert(sizeof(NFCWriteCommandData) == 0x15, "NFCWriteCommandData is an invalid size");
+#pragma pack(pop)
+
 struct NFCPollingCommandData {
     u8 enable_mifare;
     u8 unknown_1;
@@ -575,10 +602,9 @@ struct NFCPollingCommandData {
 static_assert(sizeof(NFCPollingCommandData) == 0x05, "NFCPollingCommandData is an invalid size");
 
 struct NFCRequestState {
-    MCUSubCommand sub_command;
-    NFCReadCommand command_argument;
+    NFCCommand command_argument;
+    u8 block_id;
     u8 packet_id;
-    INSERT_PADDING_BYTES(0x1);
     MCUPacketFlag packet_flag;
     u8 data_length;
     union {
@@ -587,9 +613,22 @@ struct NFCRequestState {
         NFCPollingCommandData nfc_polling;
     };
     u8 crc;
+    INSERT_PADDING_BYTES(0x1);
 };
 static_assert(sizeof(NFCRequestState) == 0x26, "NFCRequestState is an invalid size");
 
+struct NFCDataChunk {
+    u8 nfc_page;
+    u8 data_size;
+    std::array<u8, 0xFF> data;
+};
+
+struct NFCWritePackage {
+    NFCWriteCommandData command_data;
+    u8 number_of_chunks;
+    std::array<NFCDataChunk, 4> data_chunks;
+};
+
 struct IrsConfigure {
     MCUCommand command;
     MCUSubCommand sub_command;
@@ -659,7 +698,10 @@ struct SubCommandPacket {
     OutputReport output_report;
     u8 packet_counter;
     INSERT_PADDING_BYTES(0x8); // This contains vibration data
-    SubCommand sub_command;
+    union {
+        SubCommand sub_command;
+        MCUSubCommand mcu_sub_command;
+    };
     std::array<u8, 0x26> command_data;
 };
 static_assert(sizeof(SubCommandPacket) == 0x31, "SubCommandPacket is an invalid size");

src/input_common/helpers/joycon_protocol/nfc.cpp

@@ -34,6 +34,12 @@ DriverResult NfcProtocol::EnableNfc() {
 
         result = ConfigureMCU(config);
     }
+    if (result == DriverResult::Success) {
+        result = WaitSetMCUMode(ReportMode::NFC_IR_MODE_60HZ, MCUMode::NFC);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }
 
     return result;
 }
@@ -56,13 +62,20 @@ DriverResult NfcProtocol::StartNFCPollingMode() {
     LOG_DEBUG(Input, "Start NFC pooling Mode");
     ScopedSetBlocking sb(this);
     DriverResult result{DriverResult::Success};
-    TagFoundData tag_data{};
 
     if (result == DriverResult::Success) {
-        result = WaitSetMCUMode(ReportMode::NFC_IR_MODE_60HZ, MCUMode::NFC);
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
     }
     if (result == DriverResult::Success) {
-        result = WaitUntilNfcIsReady();
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStartPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Polling);
     }
     if (result == DriverResult::Success) {
         is_enabled = true;
@@ -72,49 +85,99 @@ DriverResult NfcProtocol::StartNFCPollingMode() {
 }
 
 DriverResult NfcProtocol::ScanAmiibo(std::vector<u8>& data) {
-    LOG_DEBUG(Input, "Start NFC pooling Mode");
+    if (update_counter++ < AMIIBO_UPDATE_DELAY) {
+        return DriverResult::Delayed;
+    }
+    update_counter = 0;
+
+    LOG_DEBUG(Input, "Scan for amiibos");
     ScopedSetBlocking sb(this);
     DriverResult result{DriverResult::Success};
     TagFoundData tag_data{};
 
     if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
-    }
-    if (result == DriverResult::Success) {
-        result = ReadTag(tag_data);
-    }
-    if (result == DriverResult::Success) {
-        result = WaitUntilNfcIsReady();
-    }
-    if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
+        result = IsTagInRange(tag_data);
     }
+
     if (result == DriverResult::Success) {
+        std::string uuid_string;
+        for (auto& content : tag_data.uuid) {
+            uuid_string += fmt::format(" {:02x}", content);
+        }
+        LOG_INFO(Input, "Tag detected, type={}, uuid={}", tag_data.type, uuid_string);
         result = GetAmiiboData(data);
     }
 
     return result;
 }
 
+DriverResult NfcProtocol::WriteAmiibo(std::span<const u8> data) {
+    LOG_DEBUG(Input, "Write amiibo");
+    ScopedSetBlocking sb(this);
+    DriverResult result{DriverResult::Success};
+    TagUUID tag_uuid = GetTagUUID(data);
+    TagFoundData tag_data{};
+
+    if (result == DriverResult::Success) {
+        result = IsTagInRange(tag_data, 7);
+    }
+    if (result == DriverResult::Success) {
+        if (tag_data.uuid != tag_uuid) {
+            result = DriverResult::InvalidParameters;
+        }
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::Ready);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStartPollingRequest(output, true);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::WriteReady);
+    }
+    if (result == DriverResult::Success) {
+        result = WriteAmiiboData(tag_uuid, data);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIs(NFCStatus::WriteDone);
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
+    }
+
+    return result;
+}
+
 bool NfcProtocol::HasAmiibo() {
+    if (update_counter++ < AMIIBO_UPDATE_DELAY) {
+        return true;
+    }
+    update_counter = 0;
+
     ScopedSetBlocking sb(this);
     DriverResult result{DriverResult::Success};
     TagFoundData tag_data{};
 
     if (result == DriverResult::Success) {
-        result = StartPolling(tag_data);
+        result = IsTagInRange(tag_data, 7);
     }
 
     return result == DriverResult::Success;
 }
 
-DriverResult NfcProtocol::WaitUntilNfcIsReady() {
+DriverResult NfcProtocol::WaitUntilNfcIs(NFCStatus status) {
     constexpr std::size_t timeout_limit = 10;
     MCUCommandResponse output{};
     std::size_t tries = 0;
 
     do {
-        auto result = SendStartWaitingRecieveRequest(output);
+        auto result = SendNextPackageRequest(output, {});
 
         if (result != DriverResult::Success) {
             return result;
@@ -124,19 +187,17 @@ DriverResult NfcProtocol::WaitUntilNfcIsReady() {
         }
     } while (output.mcu_report != MCUReport::NFCState ||
              (output.mcu_data[1] << 8) + output.mcu_data[0] != 0x0500 ||
-             output.mcu_data[5] != 0x31 || output.mcu_data[6] != 0x00);
+             output.mcu_data[5] != 0x31 || output.mcu_data[6] != static_cast<u8>(status));
 
     return DriverResult::Success;
 }
 
-DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
-    LOG_DEBUG(Input, "Start Polling for tag");
-    constexpr std::size_t timeout_limit = 7;
+DriverResult NfcProtocol::IsTagInRange(TagFoundData& data, std::size_t timeout_limit) {
     MCUCommandResponse output{};
     std::size_t tries = 0;
 
     do {
-        const auto result = SendStartPollingRequest(output);
+        const auto result = SendNextPackageRequest(output, {});
         if (result != DriverResult::Success) {
             return result;
         }
@@ -145,94 +206,31 @@ DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
         }
     } while (output.mcu_report != MCUReport::NFCState ||
              (output.mcu_data[1] << 8) + output.mcu_data[0] != 0x0500 ||
-             output.mcu_data[6] != 0x09);
+             (output.mcu_data[6] != 0x09 && output.mcu_data[6] != 0x04));
 
     data.type = output.mcu_data[12];
-    data.uuid.resize(output.mcu_data[14]);
+    data.uuid_size = std::min(output.mcu_data[14], static_cast<u8>(sizeof(TagUUID)));
     memcpy(data.uuid.data(), output.mcu_data.data() + 15, data.uuid.size());
 
     return DriverResult::Success;
 }
 
-DriverResult NfcProtocol::ReadTag(const TagFoundData& data) {
-    constexpr std::size_t timeout_limit = 10;
-    MCUCommandResponse output{};
-    std::size_t tries = 0;
-
-    std::string uuid_string;
-    for (auto& content : data.uuid) {
-        uuid_string += fmt::format(" {:02x}", content);
-    }
-
-    LOG_INFO(Input, "Tag detected, type={}, uuid={}", data.type, uuid_string);
-
-    tries = 0;
-    NFCPages ntag_pages = NFCPages::Block0;
-    // Read Tag data
-    while (true) {
-        auto result = SendReadAmiiboRequest(output, ntag_pages);
-        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
-
-        if (result != DriverResult::Success) {
-            return result;
-        }
-
-        if ((output.mcu_report == MCUReport::NFCReadData ||
-             output.mcu_report == MCUReport::NFCState) &&
-            nfc_status == NFCStatus::TagLost) {
-            return DriverResult::ErrorReadingData;
-        }
-
-        if (output.mcu_report == MCUReport::NFCReadData && output.mcu_data[1] == 0x07 &&
-            output.mcu_data[2] == 0x01) {
-            if (data.type != 2) {
-                continue;
-            }
-            switch (output.mcu_data[24]) {
-            case 0:
-                ntag_pages = NFCPages::Block135;
-                break;
-            case 3:
-                ntag_pages = NFCPages::Block45;
-                break;
-            case 4:
-                ntag_pages = NFCPages::Block231;
-                break;
-            default:
-                return DriverResult::ErrorReadingData;
-            }
-            continue;
-        }
-
-        if (output.mcu_report == MCUReport::NFCState && nfc_status == NFCStatus::LastPackage) {
-            // finished
-            SendStopPollingRequest(output);
-            return DriverResult::Success;
-        }
-
-        // Ignore other state reports
-        if (output.mcu_report == MCUReport::NFCState) {
-            continue;
-        }
-
-        if (tries++ > timeout_limit) {
-            return DriverResult::Timeout;
-        }
-    }
-
-    return DriverResult::Success;
-}
-
 DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
-    constexpr std::size_t timeout_limit = 10;
+    constexpr std::size_t timeout_limit = 60;
     MCUCommandResponse output{};
     std::size_t tries = 0;
 
-    NFCPages ntag_pages = NFCPages::Block135;
+    u8 package_index = 0;
     std::size_t ntag_buffer_pos = 0;
+    auto result = SendReadAmiiboRequest(output, NFCPages::Block135);
+
+    if (result != DriverResult::Success) {
+        return result;
+    }
+
     // Read Tag data
-    while (true) {
-        auto result = SendReadAmiiboRequest(output, ntag_pages);
+    while (tries++ < timeout_limit) {
+        result = SendNextPackageRequest(output, package_index);
         const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
 
         if (result != DriverResult::Success) {
@@ -255,6 +253,7 @@ DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
                 memcpy(ntag_data.data() + ntag_buffer_pos, output.mcu_data.data() + 6,
                        payload_size);
             }
+            package_index++;
             continue;
         }
 
@@ -262,32 +261,99 @@ DriverResult NfcProtocol::GetAmiiboData(std::vector<u8>& ntag_data) {
             LOG_INFO(Input, "Finished reading amiibo");
             return DriverResult::Success;
         }
+    }
 
-        // Ignore other state reports
-        if (output.mcu_report == MCUReport::NFCState) {
+    return DriverResult::Timeout;
+}
+
+DriverResult NfcProtocol::WriteAmiiboData(const TagUUID& tag_uuid, std::span<const u8> data) {
+    constexpr std::size_t timeout_limit = 60;
+    const auto nfc_data = MakeAmiiboWritePackage(tag_uuid, data);
+    const std::vector<u8> nfc_buffer_data = SerializeWritePackage(nfc_data);
+    std::span<const u8> buffer(nfc_buffer_data);
+    MCUCommandResponse output{};
+    u8 block_id = 1;
+    u8 package_index = 0;
+    std::size_t tries = 0;
+    std::size_t current_position = 0;
+
+    LOG_INFO(Input, "Writing amiibo data");
+
+    auto result = SendWriteAmiiboRequest(output, tag_uuid);
+
+    if (result != DriverResult::Success) {
+        return result;
+    }
+
+    // Read Tag data but ignore the actual sent data
+    while (tries++ < timeout_limit) {
+        result = SendNextPackageRequest(output, package_index);
+        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
+
+        if (result != DriverResult::Success) {
+            return result;
+        }
+
+        if ((output.mcu_report == MCUReport::NFCReadData ||
+             output.mcu_report == MCUReport::NFCState) &&
+            nfc_status == NFCStatus::TagLost) {
+            return DriverResult::ErrorReadingData;
+        }
+
+        if (output.mcu_report == MCUReport::NFCReadData && output.mcu_data[1] == 0x07) {
+            package_index++;
             continue;
         }
 
-        if (tries++ > timeout_limit) {
-            return DriverResult::Timeout;
+        if (output.mcu_report == MCUReport::NFCState && nfc_status == NFCStatus::LastPackage) {
+            LOG_INFO(Input, "Finished reading amiibo");
+            break;
         }
     }
 
-    return DriverResult::Success;
+    // Send Data. Nfc buffer size is 31, Send the data in smaller packages
+    while (current_position < buffer.size() && tries++ < timeout_limit) {
+        const std::size_t next_position =
+            std::min(current_position + sizeof(NFCRequestState::raw_data), buffer.size());
+        const std::size_t block_size = next_position - current_position;
+        const bool is_last_packet = block_size < sizeof(NFCRequestState::raw_data);
+
+        SendWriteDataAmiiboRequest(output, block_id, is_last_packet,
+                                   buffer.subspan(current_position, block_size));
+
+        const auto nfc_status = static_cast<NFCStatus>(output.mcu_data[6]);
+
+        if ((output.mcu_report == MCUReport::NFCReadData ||
+             output.mcu_report == MCUReport::NFCState) &&
+            nfc_status == NFCStatus::TagLost) {
+            return DriverResult::ErrorReadingData;
+        }
+
+        // Increase position when data is confirmed by the joycon
+        if (output.mcu_report == MCUReport::NFCState &&
+            (output.mcu_data[1] << 8) + output.mcu_data[0] == 0x0500 &&
+            output.mcu_data[3] == block_id) {
+            block_id++;
+            current_position = next_position;
+        }
+    }
+
+    return result;
 }
 
-DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
+DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output,
+                                                  bool is_second_attempt) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
-        .command_argument = NFCReadCommand::StartPolling,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StartPolling,
+        .block_id = {},
+        .packet_id = {},
         .packet_flag = MCUPacketFlag::LastCommandPacket,
         .data_length = sizeof(NFCPollingCommandData),
         .nfc_polling =
             {
-                .enable_mifare = 0x01,
-                .unknown_1 = 0x00,
-                .unknown_2 = 0x00,
+                .enable_mifare = 0x00,
+                .unknown_1 = static_cast<u8>(is_second_attempt ? 0xe8 : 0x00),
+                .unknown_2 = static_cast<u8>(is_second_attempt ? 0x03 : 0x00),
                 .unknown_3 = 0x2c,
                 .unknown_4 = 0x01,
             },
@@ -296,58 +362,60 @@ DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
-        .command_argument = NFCReadCommand::StopPolling,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StopPolling,
+        .block_id = {},
+        .packet_id = {},
         .packet_flag = MCUPacketFlag::LastCommandPacket,
-        .data_length = 0,
+        .data_length = {},
         .raw_data = {},
         .crc = {},
     };
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
-DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& output) {
+DriverResult NfcProtocol::SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
-        .command_argument = NFCReadCommand::StartWaitingRecieve,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::StartWaitingRecieve,
+        .block_id = {},
+        .packet_id = packet_id,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
-        .data_length = 0,
+        .data_length = {},
         .raw_data = {},
         .crc = {},
     };
 
     std::vector<u8> request_data(sizeof(NFCRequestState));
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
 }
 
 DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages) {
     NFCRequestState request{
-        .sub_command = MCUSubCommand::ReadDeviceMode,
-        .command_argument = NFCReadCommand::Ntag,
-        .packet_id = 0x0,
+        .command_argument = NFCCommand::ReadNtag,
+        .block_id = {},
+        .packet_id = {},
         .packet_flag = MCUPacketFlag::LastCommandPacket,
         .data_length = sizeof(NFCReadCommandData),
         .nfc_read =
             {
                 .unknown = 0xd0,
-                .uuid_length = 0x07,
-                .unknown_2 = 0x00,
+                .uuid_length = sizeof(NFCReadCommandData::uid),
                 .uid = {},
-                .tag_type = NFCTagType::AllTags,
+                .tag_type = NFCTagType::Ntag215,
                 .read_block = GetReadBlockCommand(ntag_pages),
             },
         .crc = {},
@@ -355,8 +423,124 @@ DriverResult NfcProtocol::SendReadAmiiboRequest(MCUCommandResponse& output, NFCP
 
     std::array<u8, sizeof(NFCRequestState)> request_data{};
     memcpy(request_data.data(), &request, sizeof(NFCRequestState));
-    request_data[37] = CalculateMCU_CRC8(request_data.data() + 1, 36);
-    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, SubCommand::STATE, request_data, output);
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
+}
+
+DriverResult NfcProtocol::SendWriteAmiiboRequest(MCUCommandResponse& output,
+                                                 const TagUUID& tag_uuid) {
+    NFCRequestState request{
+        .command_argument = NFCCommand::ReadNtag,
+        .block_id = {},
+        .packet_id = {},
+        .packet_flag = MCUPacketFlag::LastCommandPacket,
+        .data_length = sizeof(NFCReadCommandData),
+        .nfc_read =
+            {
+                .unknown = 0xd0,
+                .uuid_length = sizeof(NFCReadCommandData::uid),
+                .uid = tag_uuid,
+                .tag_type = NFCTagType::Ntag215,
+                .read_block = GetReadBlockCommand(NFCPages::Block3),
+            },
+        .crc = {},
+    };
+
+    std::array<u8, sizeof(NFCRequestState)> request_data{};
+    memcpy(request_data.data(), &request, sizeof(NFCRequestState));
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
+}
+
+DriverResult NfcProtocol::SendWriteDataAmiiboRequest(MCUCommandResponse& output, u8 block_id,
+                                                     bool is_last_packet,
+                                                     std::span<const u8> data) {
+    const auto data_size = std::min(data.size(), sizeof(NFCRequestState::raw_data));
+    NFCRequestState request{
+        .command_argument = NFCCommand::WriteNtag,
+        .block_id = block_id,
+        .packet_id = {},
+        .packet_flag =
+            is_last_packet ? MCUPacketFlag::LastCommandPacket : MCUPacketFlag::MorePacketsRemaining,
+        .data_length = static_cast<u8>(data_size),
+        .raw_data = {},
+        .crc = {},
+    };
+    memcpy(request.raw_data.data(), data.data(), data_size);
+
+    std::array<u8, sizeof(NFCRequestState)> request_data{};
+    memcpy(request_data.data(), &request, sizeof(NFCRequestState));
+    request_data[36] = CalculateMCU_CRC8(request_data.data(), 36);
+    return SendMCUData(ReportMode::NFC_IR_MODE_60HZ, MCUSubCommand::ReadDeviceMode, request_data,
+                       output);
+}
+
+std::vector<u8> NfcProtocol::SerializeWritePackage(const NFCWritePackage& package) const {
+    const std::size_t header_size =
+        sizeof(NFCWriteCommandData) + sizeof(NFCWritePackage::number_of_chunks);
+    std::vector<u8> serialized_data(header_size);
+    std::size_t start_index = 0;
+
+    memcpy(serialized_data.data(), &package, header_size);
+    start_index += header_size;
+
+    for (const auto& data_chunk : package.data_chunks) {
+        const std::size_t chunk_size =
+            sizeof(NFCDataChunk::nfc_page) + sizeof(NFCDataChunk::data_size) + data_chunk.data_size;
+
+        serialized_data.resize(start_index + chunk_size);
+        memcpy(serialized_data.data() + start_index, &data_chunk, chunk_size);
+        start_index += chunk_size;
+    }
+
+    return serialized_data;
+}
+
+NFCWritePackage NfcProtocol::MakeAmiiboWritePackage(const TagUUID& tag_uuid,
+                                                    std::span<const u8> data) const {
+    return {
+        .command_data{
+            .unknown = 0xd0,
+            .uuid_length = sizeof(NFCReadCommandData::uid),
+            .uid = tag_uuid,
+            .tag_type = NFCTagType::Ntag215,
+            .unknown2 = 0x00,
+            .unknown3 = 0x01,
+            .unknown4 = 0x04,
+            .unknown5 = 0xff,
+            .unknown6 = 0xff,
+            .unknown7 = 0xff,
+            .unknown8 = 0xff,
+            .magic = data[16],
+            .write_count = static_cast<u16>((data[17] << 8) + data[18]),
+            .amiibo_version = data[19],
+        },
+        .number_of_chunks = 3,
+        .data_chunks =
+            {
+                MakeAmiiboChunk(0x05, 0x20, data),
+                MakeAmiiboChunk(0x20, 0xf0, data),
+                MakeAmiiboChunk(0x5c, 0x98, data),
+            },
+    };
+}
+
+NFCDataChunk NfcProtocol::MakeAmiiboChunk(u8 page, u8 size, std::span<const u8> data) const {
+    constexpr u8 PAGE_SIZE = 4;
+
+    if (static_cast<std::size_t>(page * PAGE_SIZE) + size >= data.size()) {
+        return {};
+    }
+
+    NFCDataChunk chunk{
+        .nfc_page = page,
+        .data_size = size,
+        .data = {},
+    };
+    std::memcpy(chunk.data.data(), data.data() + (page * PAGE_SIZE), size);
+    return chunk;
 }
 
 NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {
@@ -365,6 +549,14 @@ NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {
         return {
             .block_count = 1,
         };
+    case NFCPages::Block3:
+        return {
+            .block_count = 1,
+            .blocks =
+                {
+                    NFCReadBlock{0x03, 0x03},
+                },
+        };
     case NFCPages::Block45:
         return {
             .block_count = 1,
@@ -399,6 +591,17 @@ NFCReadBlockCommand NfcProtocol::GetReadBlockCommand(NFCPages pages) const {
     };
 }
 
+TagUUID NfcProtocol::GetTagUUID(std::span<const u8> data) const {
+    if (data.size() < 10) {
+        return {};
+    }
+
+    // crc byte 3 is omitted in this operation
+    return {
+        data[0], data[1], data[2], data[4], data[5], data[6], data[7],
+    };
+}
+
 bool NfcProtocol::IsEnabled() const {
     return is_enabled;
 }

src/input_common/helpers/joycon_protocol/nfc.h

@@ -27,35 +27,56 @@ public:
 
     DriverResult ScanAmiibo(std::vector<u8>& data);
 
+    DriverResult WriteAmiibo(std::span<const u8> data);
+
     bool HasAmiibo();
 
     bool IsEnabled() const;
 
 private:
+    // Number of times the function will be delayed until it outputs valid data
+    static constexpr std::size_t AMIIBO_UPDATE_DELAY = 15;
+
     struct TagFoundData {
         u8 type;
-        std::vector<u8> uuid;
+        u8 uuid_size;
+        TagUUID uuid;
     };
 
-    DriverResult WaitUntilNfcIsReady();
+    DriverResult WaitUntilNfcIs(NFCStatus status);
 
-    DriverResult StartPolling(TagFoundData& data);
-
-    DriverResult ReadTag(const TagFoundData& data);
+    DriverResult IsTagInRange(TagFoundData& data, std::size_t timeout_limit = 1);
 
     DriverResult GetAmiiboData(std::vector<u8>& data);
 
-    DriverResult SendStartPollingRequest(MCUCommandResponse& output);
+    DriverResult WriteAmiiboData(const TagUUID& tag_uuid, std::span<const u8> data);
+
+    DriverResult SendStartPollingRequest(MCUCommandResponse& output,
+                                         bool is_second_attempt = false);
 
     DriverResult SendStopPollingRequest(MCUCommandResponse& output);
 
-    DriverResult SendStartWaitingRecieveRequest(MCUCommandResponse& output);
+    DriverResult SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id);
 
     DriverResult SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages);
 
+    DriverResult SendWriteAmiiboRequest(MCUCommandResponse& output, const TagUUID& tag_uuid);
+
+    DriverResult SendWriteDataAmiiboRequest(MCUCommandResponse& output, u8 block_id,
+                                            bool is_last_packet, std::span<const u8> data);
+
+    std::vector<u8> SerializeWritePackage(const NFCWritePackage& package) const;
+
+    NFCWritePackage MakeAmiiboWritePackage(const TagUUID& tag_uuid, std::span<const u8> data) const;
+
+    NFCDataChunk MakeAmiiboChunk(u8 page, u8 size, std::span<const u8> data) const;
+
     NFCReadBlockCommand GetReadBlockCommand(NFCPages pages) const;
 
+    TagUUID GetTagUUID(std::span<const u8> data) const;
+
     bool is_enabled{};
+    std::size_t update_counter{};
 };
 
 } // namespace InputCommon::Joycon

src/input_common/input_engine.cpp

@@ -380,13 +380,16 @@ void InputEngine::TriggerOnMotionChange(const PadIdentifier& identifier, int mot
     if (!configuring || !mapping_callback.on_data) {
         return;
     }
+    const auto old_value = GetMotion(identifier, motion);
     bool is_active = false;
-    if (std::abs(value.accel_x) > 1.5f || std::abs(value.accel_y) > 1.5f ||
-        std::abs(value.accel_z) > 1.5f) {
+    if (std::abs(value.accel_x - old_value.accel_x) > 1.5f ||
+        std::abs(value.accel_y - old_value.accel_y) > 1.5f ||
+        std::abs(value.accel_z - old_value.accel_z) > 1.5f) {
         is_active = true;
     }
-    if (std::abs(value.gyro_x) > 0.6f || std::abs(value.gyro_y) > 0.6f ||
-        std::abs(value.gyro_z) > 0.6f) {
+    if (std::abs(value.gyro_x - old_value.gyro_x) > 0.6f ||
+        std::abs(value.gyro_y - old_value.gyro_y) > 0.6f ||
+        std::abs(value.gyro_z - old_value.gyro_z) > 0.6f) {
         is_active = true;
     }
     if (!is_active) {

src/input_common/input_poller.cpp

@@ -667,7 +667,7 @@ public:
             .raw_value = input_engine->GetAxis(identifier, axis_z),
             .properties = properties_z,
         };
-        status.delta_timestamp = 5000;
+        status.delta_timestamp = 1000;
         status.force_update = true;
         return status;
     }

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -339,9 +339,7 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.vertex_id));
         } else {
-            const Id index{ctx.OpLoad(ctx.U32[1], ctx.vertex_index)};
-            const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
-            return ctx.OpBitcast(ctx.F32[1], ctx.OpISub(ctx.U32[1], index, base));
+            return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.vertex_index));
         }
     case IR::Attribute::BaseInstance:
         return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.base_instance));
@@ -386,9 +384,7 @@ Id EmitGetAttributeU32(EmitContext& ctx, IR::Attribute attr, Id) {
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpLoad(ctx.U32[1], ctx.vertex_id);
         } else {
-            const Id index{ctx.OpLoad(ctx.U32[1], ctx.vertex_index)};
-            const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
-            return ctx.OpISub(ctx.U32[1], index, base);
+            return ctx.OpLoad(ctx.U32[1], ctx.vertex_index);
         }
     case IR::Attribute::BaseInstance:
         return ctx.OpLoad(ctx.U32[1], ctx.base_instance);

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

@@ -102,12 +102,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
         }
         IR::F32 value{v.ir.CompositeExtract(sample, element)};
         if (element < 2) {
-            IR::U32 casted_value;
-            if (element == 0) {
-                casted_value = v.ir.ConvertFToU(32, value);
-            } else {
-                casted_value = v.ir.ConvertFToS(16, value);
-            }
+            IR::U32 casted_value = v.ir.ConvertFToU(32, value);
             v.X(dest_reg, v.ir.ShiftLeftLogical(casted_value, v.ir.Imm32(8)));
         } else {
             v.F(dest_reg, value);

src/tests/video_core/memory_tracker.cpp

@@ -535,12 +535,12 @@ TEST_CASE("MemoryTracker: Cached write downloads") {
     memory_track->MarkRegionAsGpuModified(c + PAGE, PAGE);
     int num = 0;
     memory_track->ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 1);
+    REQUIRE(num == 0);
     num = 0;
     memory_track->ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
     REQUIRE(num == 0);
     REQUIRE(!memory_track->IsRegionCpuModified(c + PAGE, PAGE));
-    REQUIRE(!memory_track->IsRegionGpuModified(c + PAGE, PAGE));
+    REQUIRE(memory_track->IsRegionGpuModified(c + PAGE, PAGE));
     memory_track->FlushCachedWrites();
     REQUIRE(memory_track->IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!memory_track->IsRegionGpuModified(c + PAGE, PAGE));

src/video_core/CMakeLists.txt

@@ -246,10 +246,14 @@ add_library(video_core STATIC
     texture_cache/util.h
     textures/astc.h
     textures/astc.cpp
+    textures/bcn.cpp
+    textures/bcn.h
     textures/decoders.cpp
     textures/decoders.h
     textures/texture.cpp
     textures/texture.h
+    textures/workers.cpp
+    textures/workers.h
     transform_feedback.cpp
     transform_feedback.h
     video_core.cpp
@@ -275,7 +279,7 @@ add_library(video_core STATIC
 create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)
-target_link_libraries(video_core PUBLIC glad shader_recompiler)
+target_link_libraries(video_core PUBLIC glad shader_recompiler stb)
 
 if (YUZU_USE_BUNDLED_FFMPEG AND NOT WIN32)
     add_dependencies(video_core ffmpeg-build)

src/video_core/buffer_cache/buffer_base.h

@@ -18,6 +18,7 @@ namespace VideoCommon {
 enum class BufferFlagBits {
     Picked = 1 << 0,
     CachedWrites = 1 << 1,
+    PreemtiveDownload = 1 << 2,
 };
 DECLARE_ENUM_FLAG_OPERATORS(BufferFlagBits)
 
@@ -54,6 +55,10 @@ public:
         flags |= BufferFlagBits::Picked;
     }
 
+    void MarkPreemtiveDownload() noexcept {
+        flags |= BufferFlagBits::PreemtiveDownload;
+    }
+
     /// Unmark buffer as picked
     void Unpick() noexcept {
         flags &= ~BufferFlagBits::Picked;
@@ -84,6 +89,10 @@ public:
         return True(flags & BufferFlagBits::CachedWrites);
     }
 
+    bool IsPreemtiveDownload() const noexcept {
+        return True(flags & BufferFlagBits::PreemtiveDownload);
+    }
+
     /// Returns the base CPU address of the buffer
     [[nodiscard]] VAddr CpuAddr() const noexcept {
         return cpu_addr;

src/video_core/buffer_cache/buffer_cache.h

@@ -23,8 +23,6 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
     common_ranges.clear();
     inline_buffer_id = NULL_BUFFER_ID;
 
-    active_async_buffers = !Settings::IsGPULevelHigh();
-
     if (!runtime.CanReportMemoryUsage()) {
         minimum_memory = DEFAULT_EXPECTED_MEMORY;
         critical_memory = DEFAULT_CRITICAL_MEMORY;
@@ -32,8 +30,8 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
     }
 
     const s64 device_memory = static_cast<s64>(runtime.GetDeviceLocalMemory());
-    const s64 min_spacing_expected = device_memory - 1_GiB - 512_MiB;
-    const s64 min_spacing_critical = device_memory - 1_GiB;
+    const s64 min_spacing_expected = device_memory - 1_GiB;
+    const s64 min_spacing_critical = device_memory - 512_MiB;
     const s64 mem_threshold = std::min(device_memory, TARGET_THRESHOLD);
     const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
     const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
@@ -75,8 +73,6 @@ void BufferCache<P>::TickFrame() {
     uniform_cache_hits[0] = 0;
     uniform_cache_shots[0] = 0;
 
-    active_async_buffers = !Settings::IsGPULevelHigh();
-
     const bool skip_preferred = hits * 256 < shots * 251;
     uniform_buffer_skip_cache_size = skip_preferred ? DEFAULT_SKIP_CACHE_SIZE : 0;
 
@@ -100,51 +96,50 @@ void BufferCache<P>::TickFrame() {
 
 template <class P>
 void BufferCache<P>::WriteMemory(VAddr cpu_addr, u64 size) {
-    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
     if (memory_tracker.IsRegionGpuModified(cpu_addr, size)) {
         const IntervalType subtract_interval{cpu_addr, cpu_addr + size};
         ClearDownload(subtract_interval);
         common_ranges.subtract(subtract_interval);
     }
+    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
 }
 
 template <class P>
 void BufferCache<P>::CachedWriteMemory(VAddr cpu_addr, u64 size) {
     memory_tracker.CachedCpuWrite(cpu_addr, size);
-    const IntervalType add_interval{Common::AlignDown(cpu_addr, YUZU_PAGESIZE),
-                                    Common::AlignUp(cpu_addr + size, YUZU_PAGESIZE)};
-    cached_ranges.add(add_interval);
+}
+
+template <class P>
+std::optional<VideoCore::RasterizerDownloadArea> BufferCache<P>::GetFlushArea(VAddr cpu_addr,
+                                                                              u64 size) {
+    std::optional<VideoCore::RasterizerDownloadArea> area{};
+    area.emplace();
+    VAddr cpu_addr_start_aligned = Common::AlignDown(cpu_addr, Core::Memory::YUZU_PAGESIZE);
+    VAddr cpu_addr_end_aligned = Common::AlignUp(cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
+    area->start_address = cpu_addr_start_aligned;
+    area->end_address = cpu_addr_end_aligned;
+    if (memory_tracker.IsRegionPreflushable(cpu_addr, size)) {
+        area->preemtive = true;
+        return area;
+    };
+    area->preemtive =
+        !IsRegionGpuModified(cpu_addr_start_aligned, cpu_addr_end_aligned - cpu_addr_start_aligned);
+    memory_tracker.MarkRegionAsPreflushable(cpu_addr_start_aligned,
+                                            cpu_addr_end_aligned - cpu_addr_start_aligned);
+    return area;
 }
 
 template <class P>
 void BufferCache<P>::DownloadMemory(VAddr cpu_addr, u64 size) {
-    WaitOnAsyncFlushes(cpu_addr, size);
     ForEachBufferInRange(cpu_addr, size, [&](BufferId, Buffer& buffer) {
         DownloadBufferMemory(buffer, cpu_addr, size);
     });
 }
 
-template <class P>
-void BufferCache<P>::WaitOnAsyncFlushes(VAddr cpu_addr, u64 size) {
-    bool must_wait = false;
-    ForEachInOverlapCounter(async_downloads, cpu_addr, size,
-                            [&](VAddr, VAddr, int) { must_wait = true; });
-    bool must_release = false;
-    ForEachInRangeSet(pending_ranges, cpu_addr, size, [&](VAddr, VAddr) { must_release = true; });
-    if (must_release) {
-        std::function<void()> tmp([]() {});
-        rasterizer.SignalFence(std::move(tmp));
-    }
-    if (must_wait || must_release) {
-        rasterizer.ReleaseFences();
-    }
-}
-
 template <class P>
 void BufferCache<P>::ClearDownload(IntervalType subtract_interval) {
     RemoveEachInOverlapCounter(async_downloads, subtract_interval, -1024);
     uncommitted_ranges.subtract(subtract_interval);
-    pending_ranges.subtract(subtract_interval);
     for (auto& interval_set : committed_ranges) {
         interval_set.subtract(subtract_interval);
     }
@@ -164,7 +159,6 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     }
 
     const IntervalType subtract_interval{*cpu_dest_address, *cpu_dest_address + amount};
-    WaitOnAsyncFlushes(*cpu_src_address, static_cast<u32>(amount));
     ClearDownload(subtract_interval);
 
     BufferId buffer_a;
@@ -192,7 +186,6 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
         const IntervalType add_interval{new_base_address, new_base_address + size};
         tmp_intervals.push_back(add_interval);
         uncommitted_ranges.add(add_interval);
-        pending_ranges.add(add_interval);
     };
     ForEachInRangeSet(common_ranges, *cpu_src_address, amount, mirror);
     // This subtraction in this order is important for overlapping copies.
@@ -205,7 +198,7 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     if (has_new_downloads) {
         memory_tracker.MarkRegionAsGpuModified(*cpu_dest_address, amount);
     }
-    std::vector<u8> tmp_buffer(amount);
+    tmp_buffer.resize(amount);
     cpu_memory.ReadBlockUnsafe(*cpu_src_address, tmp_buffer.data(), amount);
     cpu_memory.WriteBlockUnsafe(*cpu_dest_address, tmp_buffer.data(), amount);
     return true;
@@ -441,9 +434,7 @@ void BufferCache<P>::BindComputeTextureBuffer(size_t tbo_index, GPUVAddr gpu_add
 
 template <class P>
 void BufferCache<P>::FlushCachedWrites() {
-    cached_write_buffer_ids.clear();
     memory_tracker.FlushCachedWrites();
-    cached_ranges.clear();
 }
 
 template <class P>
@@ -474,15 +465,13 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
 
     if (committed_ranges.empty()) {
         if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            if (active_async_buffers) {
-                async_buffers.emplace_back(std::optional<Async_Buffer>{});
-            }
+
+            async_buffers.emplace_back(std::optional<Async_Buffer>{});
         }
         return;
     }
     MICROPROFILE_SCOPE(GPU_DownloadMemory);
 
-    pending_ranges.clear();
     auto it = committed_ranges.begin();
     while (it != committed_ranges.end()) {
         auto& current_intervals = *it;
@@ -537,64 +526,65 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
     committed_ranges.clear();
     if (downloads.empty()) {
         if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            if (active_async_buffers) {
-                async_buffers.emplace_back(std::optional<Async_Buffer>{});
-            }
+
+            async_buffers.emplace_back(std::optional<Async_Buffer>{});
         }
         return;
     }
-    if (active_async_buffers) {
-        if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
-            auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
-            boost::container::small_vector<BufferCopy, 4> normalized_copies;
-            IntervalSet new_async_range{};
-            runtime.PreCopyBarrier();
-            for (auto& [copy, buffer_id] : downloads) {
-                copy.dst_offset += download_staging.offset;
-                const std::array copies{copy};
-                BufferCopy second_copy{copy};
-                Buffer& buffer = slot_buffers[buffer_id];
-                second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
-                VAddr orig_cpu_addr = static_cast<VAddr>(second_copy.src_offset);
-                const IntervalType base_interval{orig_cpu_addr, orig_cpu_addr + copy.size};
-                async_downloads += std::make_pair(base_interval, 1);
-                runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
-                normalized_copies.push_back(second_copy);
-            }
-            runtime.PostCopyBarrier();
-            pending_downloads.emplace_back(std::move(normalized_copies));
-            async_buffers.emplace_back(download_staging);
-        } else {
+    if constexpr (IMPLEMENTS_ASYNC_DOWNLOADS) {
+        auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
+        boost::container::small_vector<BufferCopy, 4> normalized_copies;
+        IntervalSet new_async_range{};
+        runtime.PreCopyBarrier();
+        for (auto& [copy, buffer_id] : downloads) {
+            copy.dst_offset += download_staging.offset;
+            const std::array copies{copy};
+            BufferCopy second_copy{copy};
+            Buffer& buffer = slot_buffers[buffer_id];
+            second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
+            VAddr orig_cpu_addr = static_cast<VAddr>(second_copy.src_offset);
+            const IntervalType base_interval{orig_cpu_addr, orig_cpu_addr + copy.size};
+            async_downloads += std::make_pair(base_interval, 1);
+            runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
+            normalized_copies.push_back(second_copy);
+        }
+        runtime.PostCopyBarrier();
+        pending_downloads.emplace_back(std::move(normalized_copies));
+        async_buffers.emplace_back(download_staging);
+    } else {
+        if (!Settings::IsGPULevelHigh()) {
             committed_ranges.clear();
             uncommitted_ranges.clear();
-        }
-    } else {
-        if constexpr (USE_MEMORY_MAPS) {
-            auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
-            runtime.PreCopyBarrier();
-            for (auto& [copy, buffer_id] : downloads) {
-                // Have in mind the staging buffer offset for the copy
-                copy.dst_offset += download_staging.offset;
-                const std::array copies{copy};
-                runtime.CopyBuffer(download_staging.buffer, slot_buffers[buffer_id], copies, false);
-            }
-            runtime.PostCopyBarrier();
-            runtime.Finish();
-            for (const auto& [copy, buffer_id] : downloads) {
-                const Buffer& buffer = slot_buffers[buffer_id];
-                const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
-                // Undo the modified offset
-                const u64 dst_offset = copy.dst_offset - download_staging.offset;
-                const u8* read_mapped_memory = download_staging.mapped_span.data() + dst_offset;
-                cpu_memory.WriteBlockUnsafe(cpu_addr, read_mapped_memory, copy.size);
-            }
         } else {
-            const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
-            for (const auto& [copy, buffer_id] : downloads) {
-                Buffer& buffer = slot_buffers[buffer_id];
-                buffer.ImmediateDownload(copy.src_offset, immediate_buffer.subspan(0, copy.size));
-                const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
-                cpu_memory.WriteBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
+            if constexpr (USE_MEMORY_MAPS) {
+                auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
+                runtime.PreCopyBarrier();
+                for (auto& [copy, buffer_id] : downloads) {
+                    // Have in mind the staging buffer offset for the copy
+                    copy.dst_offset += download_staging.offset;
+                    const std::array copies{copy};
+                    runtime.CopyBuffer(download_staging.buffer, slot_buffers[buffer_id], copies,
+                                       false);
+                }
+                runtime.PostCopyBarrier();
+                runtime.Finish();
+                for (const auto& [copy, buffer_id] : downloads) {
+                    const Buffer& buffer = slot_buffers[buffer_id];
+                    const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
+                    // Undo the modified offset
+                    const u64 dst_offset = copy.dst_offset - download_staging.offset;
+                    const u8* read_mapped_memory = download_staging.mapped_span.data() + dst_offset;
+                    cpu_memory.WriteBlockUnsafe(cpu_addr, read_mapped_memory, copy.size);
+                }
+            } else {
+                const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
+                for (const auto& [copy, buffer_id] : downloads) {
+                    Buffer& buffer = slot_buffers[buffer_id];
+                    buffer.ImmediateDownload(copy.src_offset,
+                                             immediate_buffer.subspan(0, copy.size));
+                    const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
+                    cpu_memory.WriteBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
+                }
             }
         }
     }
@@ -1213,16 +1203,14 @@ void BufferCache<P>::UpdateComputeTextureBuffers() {
 
 template <class P>
 void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size) {
-    memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);
-
     if (memory_tracker.IsRegionCpuModified(cpu_addr, size)) {
         SynchronizeBuffer(slot_buffers[buffer_id], cpu_addr, size);
     }
+    memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);
 
     const IntervalType base_interval{cpu_addr, cpu_addr + size};
     common_ranges.add(base_interval);
     uncommitted_ranges.add(base_interval);
-    pending_ranges.add(base_interval);
 }
 
 template <class P>
@@ -1629,7 +1617,6 @@ void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
     replace(transform_feedback_buffers);
     replace(compute_uniform_buffers);
     replace(compute_storage_buffers);
-    std::erase(cached_write_buffer_ids, buffer_id);
 
     // Mark the whole buffer as CPU written to stop tracking CPU writes
     if (!do_not_mark) {
@@ -1668,15 +1655,16 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
         const bool is_nvn_cbuf = cbuf_index == 0;
         // The NVN driver buffer (index 0) is known to pack the SSBO address followed by its size.
         if (is_nvn_cbuf) {
-            return gpu_memory->Read<u32>(ssbo_addr + 8);
+            const u32 ssbo_size = gpu_memory->Read<u32>(ssbo_addr + 8);
+            if (ssbo_size != 0) {
+                return ssbo_size;
+            }
         }
         // Other titles (notably Doom Eternal) may use STG/LDG on buffer addresses in custom defined
         // cbufs, which do not store the sizes adjacent to the addresses, so use the fully
         // mapped buffer size for now.
         const u32 memory_layout_size = static_cast<u32>(gpu_memory->GetMemoryLayoutSize(gpu_addr));
-        LOG_INFO(HW_GPU, "Binding storage buffer for cbuf index {}, MemoryLayoutSize 0x{:X}",
-                 cbuf_index, memory_layout_size);
-        return memory_layout_size;
+        return std::min(memory_layout_size, static_cast<u32>(8_MiB));
     }();
     const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
     if (!cpu_addr || size == 0) {

src/video_core/buffer_cache/buffer_cache_base.h

@@ -188,6 +188,8 @@ public:
 
     void DownloadMemory(VAddr cpu_addr, u64 size);
 
+    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(VAddr cpu_addr, u64 size);
+
     bool InlineMemory(VAddr dest_address, size_t copy_size, std::span<const u8> inlined_buffer);
 
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size);
@@ -379,8 +381,6 @@ private:
 
     void RunGarbageCollector();
 
-    void WaitOnAsyncFlushes(VAddr cpu_addr, u64 size);
-
     void BindHostIndexBuffer();
 
     void BindHostVertexBuffers();
@@ -541,13 +541,10 @@ private:
                        std::array<std::array<u32, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES>, Empty>
         uniform_buffer_binding_sizes{};
 
-    std::vector<BufferId> cached_write_buffer_ids;
-
     MemoryTracker memory_tracker;
     IntervalSet uncommitted_ranges;
     IntervalSet common_ranges;
     IntervalSet cached_ranges;
-    IntervalSet pending_ranges;
     std::deque<IntervalSet> committed_ranges;
 
     // Async Buffers
@@ -572,9 +569,8 @@ private:
     u64 critical_memory = 0;
     BufferId inline_buffer_id;
 
-    bool active_async_buffers = false;
-
     std::array<BufferId, ((1ULL << 39) >> CACHING_PAGEBITS)> page_table;
+    std::vector<u8> tmp_buffer;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/memory_tracker_base.h

@@ -66,6 +66,14 @@ public:
             });
     }
 
+    /// Returns true if a region has been marked as Preflushable
+    [[nodiscard]] bool IsRegionPreflushable(VAddr query_cpu_addr, u64 query_size) noexcept {
+        return IteratePages<false>(
+            query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) {
+                return manager->template IsRegionModified<Type::Preflushable>(offset, size);
+            });
+    }
+
     /// Mark region as CPU modified, notifying the rasterizer about this change
     void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) {
         IteratePages<true>(dirty_cpu_addr, query_size,
@@ -93,6 +101,15 @@ public:
                            });
     }
 
+    /// Mark region as modified from the host GPU
+    void MarkRegionAsPreflushable(VAddr dirty_cpu_addr, u64 query_size) noexcept {
+        IteratePages<true>(dirty_cpu_addr, query_size,
+                           [](Manager* manager, u64 offset, size_t size) {
+                               manager->template ChangeRegionState<Type::Preflushable, true>(
+                                   manager->GetCpuAddr() + offset, size);
+                           });
+    }
+
     /// Unmark region as modified from the host GPU
     void UnmarkRegionAsGpuModified(VAddr dirty_cpu_addr, u64 query_size) noexcept {
         IteratePages<true>(dirty_cpu_addr, query_size,
@@ -102,6 +119,15 @@ public:
                            });
     }
 
+    /// Unmark region as modified from the host GPU
+    void UnmarkRegionAsPreflushable(VAddr dirty_cpu_addr, u64 query_size) noexcept {
+        IteratePages<true>(dirty_cpu_addr, query_size,
+                           [](Manager* manager, u64 offset, size_t size) {
+                               manager->template ChangeRegionState<Type::Preflushable, false>(
+                                   manager->GetCpuAddr() + offset, size);
+                           });
+    }
+
     /// Mark region as modified from the CPU
     /// but don't mark it as modified until FlusHCachedWrites is called.
     void CachedCpuWrite(VAddr dirty_cpu_addr, u64 query_size) {

src/video_core/buffer_cache/word_manager.h

@@ -26,6 +26,7 @@ enum class Type {
     GPU,
     CachedCPU,
     Untracked,
+    Preflushable,
 };
 
 /// Vector tracking modified pages tightly packed with small vector optimization
@@ -55,17 +56,20 @@ struct Words {
             gpu.stack.fill(0);
             cached_cpu.stack.fill(0);
             untracked.stack.fill(~u64{0});
+            preflushable.stack.fill(0);
         } else {
             // Share allocation between CPU and GPU pages and set their default values
-            u64* const alloc = new u64[num_words * 4];
+            u64* const alloc = new u64[num_words * 5];
             cpu.heap = alloc;
             gpu.heap = alloc + num_words;
             cached_cpu.heap = alloc + num_words * 2;
             untracked.heap = alloc + num_words * 3;
+            preflushable.heap = alloc + num_words * 4;
             std::fill_n(cpu.heap, num_words, ~u64{0});
             std::fill_n(gpu.heap, num_words, 0);
             std::fill_n(cached_cpu.heap, num_words, 0);
             std::fill_n(untracked.heap, num_words, ~u64{0});
+            std::fill_n(preflushable.heap, num_words, 0);
         }
         // Clean up tailing bits
         const u64 last_word_size = size_bytes % BYTES_PER_WORD;
@@ -88,13 +92,14 @@ struct Words {
         gpu = rhs.gpu;
         cached_cpu = rhs.cached_cpu;
         untracked = rhs.untracked;
+        preflushable = rhs.preflushable;
         rhs.cpu.heap = nullptr;
         return *this;
     }
 
     Words(Words&& rhs) noexcept
         : size_bytes{rhs.size_bytes}, num_words{rhs.num_words}, cpu{rhs.cpu}, gpu{rhs.gpu},
-          cached_cpu{rhs.cached_cpu}, untracked{rhs.untracked} {
+          cached_cpu{rhs.cached_cpu}, untracked{rhs.untracked}, preflushable{rhs.preflushable} {
         rhs.cpu.heap = nullptr;
     }
 
@@ -129,6 +134,8 @@ struct Words {
             return std::span<u64>(cached_cpu.Pointer(IsShort()), num_words);
         } else if constexpr (type == Type::Untracked) {
             return std::span<u64>(untracked.Pointer(IsShort()), num_words);
+        } else if constexpr (type == Type::Preflushable) {
+            return std::span<u64>(preflushable.Pointer(IsShort()), num_words);
         }
     }
 
@@ -142,6 +149,8 @@ struct Words {
             return std::span<const u64>(cached_cpu.Pointer(IsShort()), num_words);
         } else if constexpr (type == Type::Untracked) {
             return std::span<const u64>(untracked.Pointer(IsShort()), num_words);
+        } else if constexpr (type == Type::Preflushable) {
+            return std::span<const u64>(preflushable.Pointer(IsShort()), num_words);
         }
     }
 
@@ -151,6 +160,7 @@ struct Words {
     WordsArray<stack_words> gpu;
     WordsArray<stack_words> cached_cpu;
     WordsArray<stack_words> untracked;
+    WordsArray<stack_words> preflushable;
 };
 
 template <class RasterizerInterface, size_t stack_words = 1>
@@ -292,6 +302,9 @@ public:
                  (pending_pointer - pending_offset) * BYTES_PER_PAGE);
         };
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if constexpr (clear) {
                 if constexpr (type == Type::CPU || type == Type::CachedCPU) {
@@ -340,8 +353,13 @@ public:
         static_assert(type != Type::Untracked);
 
         const std::span<const u64> state_words = words.template Span<type>();
+        [[maybe_unused]] const std::span<const u64> untracked_words =
+            words.template Span<Type::Untracked>();
         bool result = false;
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if (word != 0) {
                 result = true;
@@ -362,9 +380,14 @@ public:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
         const std::span<const u64> state_words = words.template Span<type>();
+        [[maybe_unused]] const std::span<const u64> untracked_words =
+            words.template Span<Type::Untracked>();
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         IterateWords(offset, size, [&](size_t index, u64 mask) {
+            if constexpr (type == Type::GPU) {
+                mask &= ~untracked_words[index];
+            }
             const u64 word = state_words[index] & mask;
             if (word == 0) {
                 return;

src/video_core/engines/maxwell_dma.cpp

@@ -223,7 +223,7 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
     write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(src_operand.address, read_buffer.data(), src_size);
-    memory_manager.ReadBlockUnsafe(dst_operand.address, write_buffer.data(), dst_size);
+    memory_manager.ReadBlock(dst_operand.address, write_buffer.data(), dst_size);
 
     UnswizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,
                      src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
@@ -288,11 +288,7 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
     write_buffer.resize_destructive(dst_size);
 
     memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
-    if (Settings::IsGPULevelExtreme()) {
-        memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
-    } else {
-        memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size);
-    }
+    memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size);
 
     // If the input is linear and the output is tiled, swizzle the input and copy it over.
     SwizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,

src/video_core/engines/sw_blitter/blitter.cpp

@@ -5,6 +5,7 @@
 #include <cmath>
 #include <vector>
 
+#include "common/scratch_buffer.h"
 #include "video_core/engines/sw_blitter/blitter.h"
 #include "video_core/engines/sw_blitter/converter.h"
 #include "video_core/memory_manager.h"
@@ -112,11 +113,11 @@ void Bilinear(std::span<const f32> input, std::span<f32> output, size_t src_widt
 } // namespace
 
 struct SoftwareBlitEngine::BlitEngineImpl {
-    std::vector<u8> tmp_buffer;
-    std::vector<u8> src_buffer;
-    std::vector<u8> dst_buffer;
-    std::vector<f32> intermediate_src;
-    std::vector<f32> intermediate_dst;
+    Common::ScratchBuffer<u8> tmp_buffer;
+    Common::ScratchBuffer<u8> src_buffer;
+    Common::ScratchBuffer<u8> dst_buffer;
+    Common::ScratchBuffer<f32> intermediate_src;
+    Common::ScratchBuffer<f32> intermediate_dst;
     ConverterFactory converter_factory;
 };
 
@@ -158,14 +159,14 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
     const auto src_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(src.format));
     const auto dst_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(dst.format));
     const size_t src_size = get_surface_size(src, src_bytes_per_pixel);
-    impl->tmp_buffer.resize(src_size);
+    impl->tmp_buffer.resize_destructive(src_size);
     memory_manager.ReadBlock(src.Address(), impl->tmp_buffer.data(), src_size);
 
     const size_t src_copy_size = src_extent_x * src_extent_y * src_bytes_per_pixel;
 
     const size_t dst_copy_size = dst_extent_x * dst_extent_y * dst_bytes_per_pixel;
 
-    impl->src_buffer.resize(src_copy_size);
+    impl->src_buffer.resize_destructive(src_copy_size);
 
     const bool no_passthrough =
         src.format != dst.format || src_extent_x != dst_extent_x || src_extent_y != dst_extent_y;
@@ -177,8 +178,10 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
 
     const auto convertion_phase_ir = [&]() {
         auto* input_converter = impl->converter_factory.GetFormatConverter(src.format);
-        impl->intermediate_src.resize((src_copy_size / src_bytes_per_pixel) * ir_components);
-        impl->intermediate_dst.resize((dst_copy_size / dst_bytes_per_pixel) * ir_components);
+        impl->intermediate_src.resize_destructive((src_copy_size / src_bytes_per_pixel) *
+                                                  ir_components);
+        impl->intermediate_dst.resize_destructive((dst_copy_size / dst_bytes_per_pixel) *
+                                                  ir_components);
         input_converter->ConvertTo(impl->src_buffer, impl->intermediate_src);
 
         if (config.filter != Fermi2D::Filter::Bilinear) {
@@ -195,7 +198,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
 
     // Do actual Blit
 
-    impl->dst_buffer.resize(dst_copy_size);
+    impl->dst_buffer.resize_destructive(dst_copy_size);
     if (src.linear == Fermi2D::MemoryLayout::BlockLinear) {
         UnswizzleSubrect(impl->src_buffer, impl->tmp_buffer, src_bytes_per_pixel, src.width,
                          src.height, src.depth, config.src_x0, config.src_y0, src_extent_x,
@@ -218,7 +221,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
     }
 
     const size_t dst_size = get_surface_size(dst, dst_bytes_per_pixel);
-    impl->tmp_buffer.resize(dst_size);
+    impl->tmp_buffer.resize_destructive(dst_size);
     memory_manager.ReadBlock(dst.Address(), impl->tmp_buffer.data(), dst_size);
 
     if (dst.linear == Fermi2D::MemoryLayout::BlockLinear) {

src/video_core/fence_manager.h

@@ -59,6 +59,11 @@ public:
         buffer_cache.AccumulateFlushes();
     }
 
+    void SignalReference() {
+        std::function<void()> do_nothing([] {});
+        SignalFence(std::move(do_nothing));
+    }
+
     void SyncOperation(std::function<void()>&& func) {
         uncommitted_operations.emplace_back(std::move(func));
     }

src/video_core/gpu.cpp

@@ -283,6 +283,21 @@ struct GPU::Impl {
         gpu_thread.FlushRegion(addr, size);
     }
 
+    VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size) {
+        auto raster_area = rasterizer->GetFlushArea(addr, size);
+        if (raster_area.preemtive) {
+            return raster_area;
+        }
+        raster_area.preemtive = true;
+        const u64 fence = RequestSyncOperation([this, &raster_area]() {
+            rasterizer->FlushRegion(raster_area.start_address,
+                                    raster_area.end_address - raster_area.start_address);
+        });
+        gpu_thread.TickGPU();
+        WaitForSyncOperation(fence);
+        return raster_area;
+    }
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
         gpu_thread.InvalidateRegion(addr, size);
@@ -538,6 +553,10 @@ void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     impl->SwapBuffers(framebuffer);
 }
 
+VideoCore::RasterizerDownloadArea GPU::OnCPURead(VAddr addr, u64 size) {
+    return impl->OnCPURead(addr, size);
+}
+
 void GPU::FlushRegion(VAddr addr, u64 size) {
     impl->FlushRegion(addr, size);
 }

src/video_core/gpu.h

@@ -10,6 +10,7 @@
 #include "core/hle/service/nvdrv/nvdata.h"
 #include "video_core/cdma_pusher.h"
 #include "video_core/framebuffer_config.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Core {
 class System;
@@ -240,6 +241,9 @@ public:
     /// Swap buffers (render frame)
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    [[nodiscard]] VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size);
+
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     void FlushRegion(VAddr addr, u64 size);
 

src/video_core/host1x/codecs/h264.cpp

@@ -111,7 +111,7 @@ const std::vector<u8>& H264::ComposeFrame(const Host1x::NvdecCommon::NvdecRegist
     writer.WriteUe(0);
 
     writer.WriteBit(context.h264_parameter_set.entropy_coding_mode_flag != 0);
-    writer.WriteBit(false);
+    writer.WriteBit(context.h264_parameter_set.pic_order_present_flag != 0);
     writer.WriteUe(0);
     writer.WriteUe(context.h264_parameter_set.num_refidx_l0_default_active);
     writer.WriteUe(context.h264_parameter_set.num_refidx_l1_default_active);
@@ -129,7 +129,7 @@ const std::vector<u8>& H264::ComposeFrame(const Host1x::NvdecCommon::NvdecRegist
     writer.WriteBit(context.h264_parameter_set.redundant_pic_cnt_present_flag != 0);
     writer.WriteBit(context.h264_parameter_set.transform_8x8_mode_flag != 0);
 
-    writer.WriteBit(true);
+    writer.WriteBit(true); // pic_scaling_matrix_present_flag
 
     for (s32 index = 0; index < 6; index++) {
         writer.WriteBit(true);

src/video_core/query_cache.h

@@ -255,7 +255,6 @@ private:
                 if (!in_range(query)) {
                     continue;
                 }
-                rasterizer.UpdatePagesCachedCount(query.GetCpuAddr(), query.SizeInBytes(), -1);
                 AsyncJobId async_job_id = query.GetAsyncJob();
                 auto flush_result = query.Flush(async);
                 if (async_job_id == NULL_ASYNC_JOB_ID) {
@@ -273,7 +272,6 @@ private:
 
     /// Registers the passed parameters as cached and returns a pointer to the stored cached query.
     CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) {
-        rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1);
         const u64 page = static_cast<u64>(cpu_addr) >> YUZU_PAGEBITS;
         return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr,
                                                   host_ptr);

src/video_core/rasterizer_download_area.h

@@ -0,0 +1,16 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace VideoCore {
+
+struct RasterizerDownloadArea {
+    VAddr start_address;
+    VAddr end_address;
+    bool preemtive;
+};
+
+} // namespace VideoCore

src/video_core/rasterizer_interface.h

@@ -12,6 +12,7 @@
 #include "video_core/cache_types.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/gpu.h"
+#include "video_core/rasterizer_download_area.h"
 
 namespace Tegra {
 class MemoryManager;
@@ -95,6 +96,8 @@ public:
     virtual bool MustFlushRegion(VAddr addr, u64 size,
                                  VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
+    virtual RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) = 0;
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     virtual void InvalidateRegion(VAddr addr, u64 size,
                                   VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;

src/video_core/renderer_null/null_rasterizer.cpp

@@ -1,6 +1,8 @@
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/alignment.h"
+#include "core/memory.h"
 #include "video_core/host1x/host1x.h"
 #include "video_core/memory_manager.h"
 #include "video_core/renderer_null/null_rasterizer.h"
@@ -46,6 +48,14 @@ bool RasterizerNull::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheTyp
 }
 void RasterizerNull::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
 void RasterizerNull::OnCPUWrite(VAddr addr, u64 size) {}
+VideoCore::RasterizerDownloadArea RasterizerNull::GetFlushArea(VAddr addr, u64 size) {
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
 void RasterizerNull::InvalidateGPUCache() {}
 void RasterizerNull::UnmapMemory(VAddr addr, u64 size) {}
 void RasterizerNull::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) {}

src/video_core/renderer_null/null_rasterizer.h

@@ -54,6 +54,7 @@ public:
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void OnCPUWrite(VAddr addr, u64 size) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
     void UnmapMemory(VAddr addr, u64 size) override;
     void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -433,6 +433,29 @@ bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
+VideoCore::RasterizerDownloadArea RasterizerOpenGL::GetFlushArea(VAddr addr, u64 size) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        auto area = texture_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        auto area = buffer_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
+
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
@@ -1281,7 +1304,7 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
                                        const Tegra::DMA::BufferOperand& buffer_operand,
                                        const Tegra::DMA::ImageOperand& image_operand) {
     std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
-    const auto image_id = texture_cache.DmaImageId(image_operand);
+    const auto image_id = texture_cache.DmaImageId(image_operand, IS_IMAGE_UPLOAD);
     if (image_id == VideoCommon::NULL_IMAGE_ID) {
         return false;
     }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -95,6 +95,7 @@ public:
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     bool MustFlushRegion(VAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void OnCPUWrite(VAddr addr, u64 size) override;

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -231,8 +231,10 @@ void ApplySwizzle(GLuint handle, PixelFormat format, std::array<SwizzleSource, 4
 
 [[nodiscard]] bool CanBeAccelerated(const TextureCacheRuntime& runtime,
                                     const VideoCommon::ImageInfo& info) {
-    if (IsPixelFormatASTC(info.format) && !runtime.HasNativeASTC()) {
+    if (IsPixelFormatASTC(info.format) && info.size.depth == 1 && !runtime.HasNativeASTC()) {
         return Settings::values.accelerate_astc.GetValue() &&
+               Settings::values.astc_recompression.GetValue() ==
+                   Settings::AstcRecompression::Uncompressed &&
                !Settings::values.async_astc.GetValue();
     }
     // Disable other accelerated uploads for now as they don't implement swizzled uploads
@@ -437,6 +439,19 @@ OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_form
     return GL_R32UI;
 }
 
+[[nodiscard]] GLenum SelectAstcFormat(PixelFormat format, bool is_srgb) {
+    switch (Settings::values.astc_recompression.GetValue()) {
+    case Settings::AstcRecompression::Bc1:
+        return is_srgb ? GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT : GL_COMPRESSED_RGBA_S3TC_DXT1_EXT;
+        break;
+    case Settings::AstcRecompression::Bc3:
+        return is_srgb ? GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT : GL_COMPRESSED_RGBA_S3TC_DXT5_EXT;
+        break;
+    default:
+        return is_srgb ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+    }
+}
+
 } // Anonymous namespace
 
 ImageBufferMap::~ImageBufferMap() {
@@ -739,9 +754,16 @@ Image::Image(TextureCacheRuntime& runtime_, const VideoCommon::ImageInfo& info_,
     if (IsConverted(runtime->device, info.format, info.type)) {
         flags |= ImageFlagBits::Converted;
         flags |= ImageFlagBits::CostlyLoad;
-        gl_internal_format = IsPixelFormatSRGB(info.format) ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+
+        const bool is_srgb = IsPixelFormatSRGB(info.format);
+        gl_internal_format = is_srgb ? GL_SRGB8_ALPHA8 : GL_RGBA8;
         gl_format = GL_RGBA;
         gl_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+        if (IsPixelFormatASTC(info.format)) {
+            gl_internal_format = SelectAstcFormat(info.format, is_srgb);
+            gl_format = GL_NONE;
+        }
     } else {
         const auto& tuple = MaxwellToGL::GetFormatTuple(info.format);
         gl_internal_format = tuple.internal_format;
@@ -1130,7 +1152,12 @@ ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewI
       views{runtime.null_image_views} {
     const Device& device = runtime.device;
     if (True(image.flags & ImageFlagBits::Converted)) {
-        internal_format = IsPixelFormatSRGB(info.format) ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+        const bool is_srgb = IsPixelFormatSRGB(info.format);
+        internal_format = is_srgb ? GL_SRGB8_ALPHA8 : GL_RGBA8;
+
+        if (IsPixelFormatASTC(info.format)) {
+            internal_format = SelectAstcFormat(info.format, is_srgb);
+        }
     } else {
         internal_format = MaxwellToGL::GetFormatTuple(format).internal_format;
     }

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -144,6 +144,10 @@ public:
         return state_tracker;
     }
 
+    void BarrierFeedbackLoop() const noexcept {
+        // OpenGL does not require a barrier for attachment feedback loops.
+    }
+
 private:
     struct StagingBuffers {
         explicit StagingBuffers(GLenum storage_flags_, GLenum map_flags_);

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -6,6 +6,7 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "common/settings.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/maxwell_to_vk.h"
 #include "video_core/surface.h"
@@ -237,14 +238,25 @@ FormatInfo SurfaceFormat(const Device& device, FormatType format_type, bool with
                          PixelFormat pixel_format) {
     ASSERT(static_cast<size_t>(pixel_format) < std::size(tex_format_tuples));
     FormatTuple tuple = tex_format_tuples[static_cast<size_t>(pixel_format)];
-    // Use A8B8G8R8_UNORM on hardware that doesn't support ASTC natively
+    // Transcode on hardware that doesn't support ASTC natively
     if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) {
         const bool is_srgb = with_srgb && VideoCore::Surface::IsPixelFormatSRGB(pixel_format);
-        if (is_srgb) {
-            tuple.format = VK_FORMAT_A8B8G8R8_SRGB_PACK32;
-        } else {
-            tuple.format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
-            tuple.usage |= Storage;
+
+        switch (Settings::values.astc_recompression.GetValue()) {
+        case Settings::AstcRecompression::Uncompressed:
+            if (is_srgb) {
+                tuple.format = VK_FORMAT_A8B8G8R8_SRGB_PACK32;
+            } else {
+                tuple.format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
+                tuple.usage |= Storage;
+            }
+            break;
+        case Settings::AstcRecompression::Bc1:
+            tuple.format = is_srgb ? VK_FORMAT_BC1_RGBA_SRGB_BLOCK : VK_FORMAT_BC1_RGBA_UNORM_BLOCK;
+            break;
+        case Settings::AstcRecompression::Bc3:
+            tuple.format = is_srgb ? VK_FORMAT_BC3_SRGB_BLOCK : VK_FORMAT_BC3_UNORM_BLOCK;
+            break;
         }
     }
     const bool attachable = (tuple.usage & Attachable) != 0;

src/video_core/renderer_vulkan/pipeline_helper.h

@@ -176,7 +176,7 @@ public:
 };
 
 inline void PushImageDescriptors(TextureCache& texture_cache,
-                                 UpdateDescriptorQueue& update_descriptor_queue,
+                                 GuestDescriptorQueue& guest_descriptor_queue,
                                  const Shader::Info& info, RescalingPushConstant& rescaling,
                                  const VkSampler*& samplers,
                                  const VideoCommon::ImageViewInOut*& views) {
@@ -190,7 +190,7 @@ inline void PushImageDescriptors(TextureCache& texture_cache,
             const VkSampler sampler{*(samplers++)};
             ImageView& image_view{texture_cache.GetImageView(image_view_id)};
             const VkImageView vk_image_view{image_view.Handle(desc.type)};
-            update_descriptor_queue.AddSampledImage(vk_image_view, sampler);
+            guest_descriptor_queue.AddSampledImage(vk_image_view, sampler);
             rescaling.PushTexture(texture_cache.IsRescaling(image_view));
         }
     }
@@ -201,7 +201,7 @@ inline void PushImageDescriptors(TextureCache& texture_cache,
                 texture_cache.MarkModification(image_view.image_id);
             }
             const VkImageView vk_image_view{image_view.StorageView(desc.type, desc.format)};
-            update_descriptor_queue.AddImage(vk_image_view);
+            guest_descriptor_queue.AddImage(vk_image_view);
             rescaling.PushImage(texture_cache.IsRescaling(image_view));
         }
     }

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -298,12 +298,14 @@ private:
 
 BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& memory_allocator_,
                                        Scheduler& scheduler_, StagingBufferPool& staging_pool_,
-                                       UpdateDescriptorQueue& update_descriptor_queue_,
+                                       GuestDescriptorQueue& guest_descriptor_queue_,
+                                       ComputePassDescriptorQueue& compute_pass_descriptor_queue,
                                        DescriptorPool& descriptor_pool)
     : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
-      staging_pool{staging_pool_}, update_descriptor_queue{update_descriptor_queue_},
-      uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
-      quad_index_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue) {
+      staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
+      uint8_pass(device, scheduler, descriptor_pool, staging_pool, compute_pass_descriptor_queue),
+      quad_index_pass(device, scheduler, descriptor_pool, staging_pool,
+                      compute_pass_descriptor_queue) {
     quad_array_index_buffer = std::make_shared<QuadArrayIndexBuffer>(device_, memory_allocator_,
                                                                      scheduler_, staging_pool_);
     quad_strip_index_buffer = std::make_shared<QuadStripIndexBuffer>(device_, memory_allocator_,

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -63,7 +63,8 @@ class BufferCacheRuntime {
 public:
     explicit BufferCacheRuntime(const Device& device_, MemoryAllocator& memory_manager_,
                                 Scheduler& scheduler_, StagingBufferPool& staging_pool_,
-                                UpdateDescriptorQueue& update_descriptor_queue_,
+                                GuestDescriptorQueue& guest_descriptor_queue,
+                                ComputePassDescriptorQueue& compute_pass_descriptor_queue,
                                 DescriptorPool& descriptor_pool);
 
     void Finish();
@@ -116,12 +117,12 @@ public:
 
     void BindTextureBuffer(Buffer& buffer, u32 offset, u32 size,
                            VideoCore::Surface::PixelFormat format) {
-        update_descriptor_queue.AddTexelBuffer(buffer.View(offset, size, format));
+        guest_descriptor_queue.AddTexelBuffer(buffer.View(offset, size, format));
     }
 
 private:
     void BindBuffer(VkBuffer buffer, u32 offset, u32 size) {
-        update_descriptor_queue.AddBuffer(buffer, offset, size);
+        guest_descriptor_queue.AddBuffer(buffer, offset, size);
     }
 
     void ReserveNullBuffer();
@@ -130,7 +131,7 @@ private:
     MemoryAllocator& memory_allocator;
     Scheduler& scheduler;
     StagingBufferPool& staging_pool;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    GuestDescriptorQueue& guest_descriptor_queue;
 
     std::shared_ptr<QuadArrayIndexBuffer> quad_array_index_buffer;
     std::shared_ptr<QuadStripIndexBuffer> quad_strip_index_buffer;

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -200,12 +200,12 @@ ComputePass::~ComputePass() = default;
 
 Uint8Pass::Uint8Pass(const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool,
                      StagingBufferPool& staging_buffer_pool_,
-                     UpdateDescriptorQueue& update_descriptor_queue_)
+                     ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
     : ComputePass(device_, descriptor_pool, INPUT_OUTPUT_DESCRIPTOR_SET_BINDINGS,
                   INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO, {},
                   VULKAN_UINT8_COMP_SPV),
       scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
-      update_descriptor_queue{update_descriptor_queue_} {}
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
 
 Uint8Pass::~Uint8Pass() = default;
 
@@ -214,10 +214,10 @@ std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer
     const u32 staging_size = static_cast<u32>(num_vertices * sizeof(u16));
     const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
 
-    update_descriptor_queue.Acquire();
-    update_descriptor_queue.AddBuffer(src_buffer, src_offset, num_vertices);
-    update_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
-    const void* const descriptor_data{update_descriptor_queue.UpdateData()};
+    compute_pass_descriptor_queue.Acquire();
+    compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, num_vertices);
+    compute_pass_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
+    const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([this, descriptor_data, num_vertices](vk::CommandBuffer cmdbuf) {
@@ -242,12 +242,12 @@ std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer
 QuadIndexedPass::QuadIndexedPass(const Device& device_, Scheduler& scheduler_,
                                  DescriptorPool& descriptor_pool_,
                                  StagingBufferPool& staging_buffer_pool_,
-                                 UpdateDescriptorQueue& update_descriptor_queue_)
+                                 ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
     : ComputePass(device_, descriptor_pool_, INPUT_OUTPUT_DESCRIPTOR_SET_BINDINGS,
                   INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO,
                   COMPUTE_PUSH_CONSTANT_RANGE<sizeof(u32) * 3>, VULKAN_QUAD_INDEXED_COMP_SPV),
       scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
-      update_descriptor_queue{update_descriptor_queue_} {}
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
 
 QuadIndexedPass::~QuadIndexedPass() = default;
 
@@ -272,10 +272,10 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
     const std::size_t staging_size = num_tri_vertices * sizeof(u32);
     const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
 
-    update_descriptor_queue.Acquire();
-    update_descriptor_queue.AddBuffer(src_buffer, src_offset, input_size);
-    update_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
-    const void* const descriptor_data{update_descriptor_queue.UpdateData()};
+    compute_pass_descriptor_queue.Acquire();
+    compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, input_size);
+    compute_pass_descriptor_queue.AddBuffer(staging.buffer, staging.offset, staging_size);
+    const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
     scheduler.RequestOutsideRenderPassOperationContext();
     scheduler.Record([this, descriptor_data, num_tri_vertices, base_vertex, index_shift,
@@ -304,13 +304,14 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
 ASTCDecoderPass::ASTCDecoderPass(const Device& device_, Scheduler& scheduler_,
                                  DescriptorPool& descriptor_pool_,
                                  StagingBufferPool& staging_buffer_pool_,
-                                 UpdateDescriptorQueue& update_descriptor_queue_,
+                                 ComputePassDescriptorQueue& compute_pass_descriptor_queue_,
                                  MemoryAllocator& memory_allocator_)
     : ComputePass(device_, descriptor_pool_, ASTC_DESCRIPTOR_SET_BINDINGS,
                   ASTC_PASS_DESCRIPTOR_UPDATE_TEMPLATE_ENTRY, ASTC_BANK_INFO,
                   COMPUTE_PUSH_CONSTANT_RANGE<sizeof(AstcPushConstants)>, ASTC_DECODER_COMP_SPV),
       scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
-      update_descriptor_queue{update_descriptor_queue_}, memory_allocator{memory_allocator_} {}
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_}, memory_allocator{
+                                                                         memory_allocator_} {}
 
 ASTCDecoderPass::~ASTCDecoderPass() = default;
 
@@ -358,11 +359,11 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         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.AddImage(image.StorageImageView(swizzle.level));
-        const void* const descriptor_data{update_descriptor_queue.UpdateData()};
+        compute_pass_descriptor_queue.Acquire();
+        compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
+                                                image.guest_size_bytes - swizzle.buffer_offset);
+        compute_pass_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
+        const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
 
         // To unswizzle the ASTC data
         const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -9,6 +9,7 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -21,7 +22,6 @@ namespace Vulkan {
 class Device;
 class StagingBufferPool;
 class Scheduler;
-class UpdateDescriptorQueue;
 class Image;
 struct StagingBufferRef;
 
@@ -50,7 +50,7 @@ class Uint8Pass final : public ComputePass {
 public:
     explicit Uint8Pass(const Device& device_, Scheduler& scheduler_,
                        DescriptorPool& descriptor_pool_, StagingBufferPool& staging_buffer_pool_,
-                       UpdateDescriptorQueue& update_descriptor_queue_);
+                       ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
     ~Uint8Pass();
 
     /// Assemble uint8 indices into an uint16 index buffer
@@ -61,7 +61,7 @@ public:
 private:
     Scheduler& scheduler;
     StagingBufferPool& staging_buffer_pool;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
 };
 
 class QuadIndexedPass final : public ComputePass {
@@ -69,7 +69,7 @@ public:
     explicit QuadIndexedPass(const Device& device_, Scheduler& scheduler_,
                              DescriptorPool& descriptor_pool_,
                              StagingBufferPool& staging_buffer_pool_,
-                             UpdateDescriptorQueue& update_descriptor_queue_);
+                             ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
     ~QuadIndexedPass();
 
     std::pair<VkBuffer, VkDeviceSize> Assemble(
@@ -79,7 +79,7 @@ public:
 private:
     Scheduler& scheduler;
     StagingBufferPool& staging_buffer_pool;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
 };
 
 class ASTCDecoderPass final : public ComputePass {
@@ -87,7 +87,7 @@ public:
     explicit ASTCDecoderPass(const Device& device_, Scheduler& scheduler_,
                              DescriptorPool& descriptor_pool_,
                              StagingBufferPool& staging_buffer_pool_,
-                             UpdateDescriptorQueue& update_descriptor_queue_,
+                             ComputePassDescriptorQueue& compute_pass_descriptor_queue_,
                              MemoryAllocator& memory_allocator_);
     ~ASTCDecoderPass();
 
@@ -97,7 +97,7 @@ public:
 private:
     Scheduler& scheduler;
     StagingBufferPool& staging_buffer_pool;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
     MemoryAllocator& memory_allocator;
 };
 

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -26,13 +26,13 @@ using Tegra::Texture::TexturePair;
 
 ComputePipeline::ComputePipeline(const Device& device_, vk::PipelineCache& pipeline_cache_,
                                  DescriptorPool& descriptor_pool,
-                                 UpdateDescriptorQueue& update_descriptor_queue_,
+                                 GuestDescriptorQueue& guest_descriptor_queue_,
                                  Common::ThreadWorker* thread_worker,
                                  PipelineStatistics* pipeline_statistics,
                                  VideoCore::ShaderNotify* shader_notify, const Shader::Info& info_,
                                  vk::ShaderModule spv_module_)
-    : device{device_}, pipeline_cache(pipeline_cache_),
-      update_descriptor_queue{update_descriptor_queue_}, info{info_},
+    : device{device_},
+      pipeline_cache(pipeline_cache_), guest_descriptor_queue{guest_descriptor_queue_}, info{info_},
       spv_module(std::move(spv_module_)) {
     if (shader_notify) {
         shader_notify->MarkShaderBuilding();
@@ -99,7 +99,7 @@ ComputePipeline::ComputePipeline(const Device& device_, vk::PipelineCache& pipel
 void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
                                 Tegra::MemoryManager& gpu_memory, Scheduler& scheduler,
                                 BufferCache& buffer_cache, TextureCache& texture_cache) {
-    update_descriptor_queue.Acquire();
+    guest_descriptor_queue.Acquire();
 
     buffer_cache.SetComputeUniformBufferState(info.constant_buffer_mask, &uniform_buffer_sizes);
     buffer_cache.UnbindComputeStorageBuffers();
@@ -194,7 +194,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
     RescalingPushConstant rescaling;
     const VkSampler* samplers_it{samplers.data()};
     const VideoCommon::ImageViewInOut* views_it{views.data()};
-    PushImageDescriptors(texture_cache, update_descriptor_queue, info, rescaling, samplers_it,
+    PushImageDescriptors(texture_cache, guest_descriptor_queue, info, rescaling, samplers_it,
                          views_it);
 
     if (!is_built.load(std::memory_order::relaxed)) {
@@ -204,7 +204,7 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
             build_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); });
         });
     }
-    const void* const descriptor_data{update_descriptor_queue.UpdateData()};
+    const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
     const bool is_rescaling = !info.texture_descriptors.empty() || !info.image_descriptors.empty();
     scheduler.Record([this, descriptor_data, is_rescaling,
                       rescaling_data = rescaling.Data()](vk::CommandBuffer cmdbuf) {

src/video_core/renderer_vulkan/vk_compute_pipeline.h

@@ -30,7 +30,7 @@ class ComputePipeline {
 public:
     explicit ComputePipeline(const Device& device, vk::PipelineCache& pipeline_cache,
                              DescriptorPool& descriptor_pool,
-                             UpdateDescriptorQueue& update_descriptor_queue,
+                             GuestDescriptorQueue& guest_descriptor_queue,
                              Common::ThreadWorker* thread_worker,
                              PipelineStatistics* pipeline_statistics,
                              VideoCore::ShaderNotify* shader_notify, const Shader::Info& info,
@@ -48,7 +48,7 @@ public:
 private:
     const Device& device;
     vk::PipelineCache& pipeline_cache;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    GuestDescriptorQueue& guest_descriptor_queue;
     Shader::Info info;
 
     VideoCommon::ComputeUniformBufferSizes uniform_buffer_sizes{};

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -236,13 +236,13 @@ GraphicsPipeline::GraphicsPipeline(
     Scheduler& scheduler_, BufferCache& buffer_cache_, TextureCache& texture_cache_,
     vk::PipelineCache& pipeline_cache_, VideoCore::ShaderNotify* shader_notify,
     const Device& device_, DescriptorPool& descriptor_pool,
-    UpdateDescriptorQueue& update_descriptor_queue_, Common::ThreadWorker* worker_thread,
+    GuestDescriptorQueue& guest_descriptor_queue_, Common::ThreadWorker* worker_thread,
     PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache,
     const GraphicsPipelineCacheKey& key_, std::array<vk::ShaderModule, NUM_STAGES> stages,
     const std::array<const Shader::Info*, NUM_STAGES>& infos)
     : key{key_}, device{device_}, texture_cache{texture_cache_}, buffer_cache{buffer_cache_},
       pipeline_cache(pipeline_cache_), scheduler{scheduler_},
-      update_descriptor_queue{update_descriptor_queue_}, spv_modules{std::move(stages)} {
+      guest_descriptor_queue{guest_descriptor_queue_}, spv_modules{std::move(stages)} {
     if (shader_notify) {
         shader_notify->MarkShaderBuilding();
     }
@@ -449,7 +449,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
     buffer_cache.UpdateGraphicsBuffers(is_indexed);
     buffer_cache.BindHostGeometryBuffers(is_indexed);
 
-    update_descriptor_queue.Acquire();
+    guest_descriptor_queue.Acquire();
 
     RescalingPushConstant rescaling;
     RenderAreaPushConstant render_area;
@@ -457,7 +457,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
     const VideoCommon::ImageViewInOut* views_it{views.data()};
     const auto prepare_stage{[&](size_t stage) LAMBDA_FORCEINLINE {
         buffer_cache.BindHostStageBuffers(stage);
-        PushImageDescriptors(texture_cache, update_descriptor_queue, stage_infos[stage], rescaling,
+        PushImageDescriptors(texture_cache, guest_descriptor_queue, stage_infos[stage], rescaling,
                              samplers_it, views_it);
         const auto& info{stage_infos[0]};
         if (info.uses_render_area) {
@@ -481,12 +481,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
     if constexpr (Spec::enabled_stages[4]) {
         prepare_stage(4);
     }
+    texture_cache.UpdateRenderTargets(false);
+    texture_cache.CheckFeedbackLoop(views);
     ConfigureDraw(rescaling, render_area);
 }
 
 void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
                                      const RenderAreaPushConstant& render_area) {
-    texture_cache.UpdateRenderTargets(false);
     scheduler.RequestRenderpass(texture_cache.GetFramebuffer());
 
     if (!is_built.load(std::memory_order::relaxed)) {
@@ -499,7 +500,7 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
     const bool is_rescaling{texture_cache.IsRescaling()};
     const bool update_rescaling{scheduler.UpdateRescaling(is_rescaling)};
     const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
-    const void* const descriptor_data{update_descriptor_queue.UpdateData()};
+    const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
     scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
                       is_rescaling, update_rescaling,
                       uses_render_area = render_area.uses_render_area,

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -64,7 +64,6 @@ class RenderPassCache;
 class RescalingPushConstant;
 class RenderAreaPushConstant;
 class Scheduler;
-class UpdateDescriptorQueue;
 
 class GraphicsPipeline {
     static constexpr size_t NUM_STAGES = Tegra::Engines::Maxwell3D::Regs::MaxShaderStage;
@@ -74,7 +73,7 @@ public:
         Scheduler& scheduler, BufferCache& buffer_cache, TextureCache& texture_cache,
         vk::PipelineCache& pipeline_cache, VideoCore::ShaderNotify* shader_notify,
         const Device& device, DescriptorPool& descriptor_pool,
-        UpdateDescriptorQueue& update_descriptor_queue, Common::ThreadWorker* worker_thread,
+        GuestDescriptorQueue& guest_descriptor_queue, Common::ThreadWorker* worker_thread,
         PipelineStatistics* pipeline_statistics, RenderPassCache& render_pass_cache,
         const GraphicsPipelineCacheKey& key, std::array<vk::ShaderModule, NUM_STAGES> stages,
         const std::array<const Shader::Info*, NUM_STAGES>& infos);
@@ -133,7 +132,7 @@ private:
     BufferCache& buffer_cache;
     vk::PipelineCache& pipeline_cache;
     Scheduler& scheduler;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    GuestDescriptorQueue& guest_descriptor_queue;
 
     void (*configure_func)(GraphicsPipeline*, bool){};
 

src/video_core/renderer_vulkan/vk_master_semaphore.cpp

@@ -10,11 +10,16 @@
 
 namespace Vulkan {
 
+constexpr u64 FENCE_RESERVE_SIZE = 8;
+
 MasterSemaphore::MasterSemaphore(const Device& device_) : device(device_) {
     if (!device.HasTimelineSemaphore()) {
         static constexpr VkFenceCreateInfo fence_ci{
             .sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO, .pNext = nullptr, .flags = 0};
-        fence = device.GetLogical().CreateFence(fence_ci);
+        free_queue.resize(FENCE_RESERVE_SIZE);
+        std::ranges::generate(free_queue,
+                              [&] { return device.GetLogical().CreateFence(fence_ci); });
+        wait_thread = std::jthread([this](std::stop_token token) { WaitThread(token); });
         return;
     }
 
@@ -167,16 +172,53 @@ VkResult MasterSemaphore::SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphor
         .pSignalSemaphores = &signal_semaphore,
     };
 
+    auto fence = GetFreeFence();
     auto result = device.GetGraphicsQueue().Submit(submit_info, *fence);
 
     if (result == VK_SUCCESS) {
-        fence.Wait();
-        fence.Reset();
-        gpu_tick.store(host_tick);
-        gpu_tick.notify_all();
+        std::scoped_lock lock{wait_mutex};
+        wait_queue.emplace(host_tick, std::move(fence));
+        wait_cv.notify_one();
     }
 
     return result;
 }
 
+void MasterSemaphore::WaitThread(std::stop_token token) {
+    while (!token.stop_requested()) {
+        u64 host_tick;
+        vk::Fence fence;
+        {
+            std::unique_lock lock{wait_mutex};
+            Common::CondvarWait(wait_cv, lock, token, [this] { return !wait_queue.empty(); });
+            if (token.stop_requested()) {
+                return;
+            }
+            std::tie(host_tick, fence) = std::move(wait_queue.front());
+            wait_queue.pop();
+        }
+
+        fence.Wait();
+        fence.Reset();
+        gpu_tick.store(host_tick);
+        gpu_tick.notify_all();
+
+        std::scoped_lock lock{free_mutex};
+        free_queue.push_front(std::move(fence));
+    }
+}
+
+vk::Fence MasterSemaphore::GetFreeFence() {
+    std::scoped_lock lock{free_mutex};
+    if (free_queue.empty()) {
+        static constexpr VkFenceCreateInfo fence_ci{
+            .sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO, .pNext = nullptr, .flags = 0};
+        return device.GetLogical().CreateFence(fence_ci);
+    }
+
+    auto fence = std::move(free_queue.back());
+    free_queue.pop_back();
+    return fence;
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_master_semaphore.h

@@ -5,8 +5,10 @@
 
 #include <atomic>
 #include <condition_variable>
+#include <deque>
 #include <mutex>
 #include <thread>
+#include <queue>
 
 #include "common/common_types.h"
 #include "common/polyfill_thread.h"
@@ -17,6 +19,8 @@ namespace Vulkan {
 class Device;
 
 class MasterSemaphore {
+    using Waitable = std::pair<u64, vk::Fence>;
+
 public:
     explicit MasterSemaphore(const Device& device);
     ~MasterSemaphore();
@@ -57,13 +61,22 @@ private:
     VkResult SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
                               VkSemaphore wait_semaphore, u64 host_tick);
 
+    void WaitThread(std::stop_token token);
+
+    vk::Fence GetFreeFence();
+
 private:
     const Device& device;             ///< Device.
-    vk::Fence fence;                  ///< Fence.
     vk::Semaphore semaphore;          ///< Timeline semaphore.
     std::atomic<u64> gpu_tick{0};     ///< Current known GPU tick.
     std::atomic<u64> current_tick{1}; ///< Current logical tick.
+    std::mutex wait_mutex;
+    std::mutex free_mutex;
+    std::condition_variable_any wait_cv;
+    std::queue<Waitable> wait_queue;  ///< Queue for the fences to be waited on by the wait thread.
+    std::deque<vk::Fence> free_queue; ///< Holds available fences for submission.
     std::jthread debug_thread;        ///< Debug thread to workaround validation layer bugs.
+    std::jthread wait_thread;         ///< Helper thread that waits for submitted fences.
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -277,11 +277,11 @@ bool GraphicsPipelineCacheKey::operator==(const GraphicsPipelineCacheKey& rhs) c
 
 PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device_,
                              Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-                             UpdateDescriptorQueue& update_descriptor_queue_,
+                             GuestDescriptorQueue& guest_descriptor_queue_,
                              RenderPassCache& render_pass_cache_, BufferCache& buffer_cache_,
                              TextureCache& texture_cache_, VideoCore::ShaderNotify& shader_notify_)
     : VideoCommon::ShaderCache{rasterizer_}, device{device_}, scheduler{scheduler_},
-      descriptor_pool{descriptor_pool_}, update_descriptor_queue{update_descriptor_queue_},
+      descriptor_pool{descriptor_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
       render_pass_cache{render_pass_cache_}, buffer_cache{buffer_cache_},
       texture_cache{texture_cache_}, shader_notify{shader_notify_},
       use_asynchronous_shaders{Settings::values.use_asynchronous_shaders.GetValue()},
@@ -643,7 +643,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
     Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr};
     return std::make_unique<GraphicsPipeline>(
         scheduler, buffer_cache, texture_cache, vulkan_pipeline_cache, &shader_notify, device,
-        descriptor_pool, update_descriptor_queue, thread_worker, statistics, render_pass_cache, key,
+        descriptor_pool, guest_descriptor_queue, thread_worker, statistics, render_pass_cache, key,
         std::move(modules), infos);
 
 } catch (const Shader::Exception& exception) {
@@ -698,7 +698,8 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
     PipelineStatistics* statistics, bool build_in_parallel) try {
     // TODO: Remove this when Intel fixes their shader compiler.
     //       https://github.com/IGCIT/Intel-GPU-Community-Issue-Tracker-IGCIT/issues/159
-    if (device.GetDriverID() == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS) {
+    if (device.GetDriverID() == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS &&
+        !Settings::values.enable_compute_pipelines.GetValue()) {
         LOG_ERROR(Render_Vulkan, "Skipping 0x{:016x}", key.Hash());
         return nullptr;
     }
@@ -722,7 +723,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
     }
     Common::ThreadWorker* const thread_worker{build_in_parallel ? &workers : nullptr};
     return std::make_unique<ComputePipeline>(device, vulkan_pipeline_cache, descriptor_pool,
-                                             update_descriptor_queue, thread_worker, statistics,
+                                             guest_descriptor_queue, thread_worker, statistics,
                                              &shader_notify, program.info, std::move(spv_module));
 
 } catch (const Shader::Exception& exception) {

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -82,7 +82,6 @@ class PipelineStatistics;
 class RasterizerVulkan;
 class RenderPassCache;
 class Scheduler;
-class UpdateDescriptorQueue;
 
 using VideoCommon::ShaderInfo;
 
@@ -102,7 +101,7 @@ class PipelineCache : public VideoCommon::ShaderCache {
 public:
     explicit PipelineCache(RasterizerVulkan& rasterizer, const Device& device, Scheduler& scheduler,
                            DescriptorPool& descriptor_pool,
-                           UpdateDescriptorQueue& update_descriptor_queue,
+                           GuestDescriptorQueue& guest_descriptor_queue,
                            RenderPassCache& render_pass_cache, BufferCache& buffer_cache,
                            TextureCache& texture_cache, VideoCore::ShaderNotify& shader_notify_);
     ~PipelineCache();
@@ -144,7 +143,7 @@ private:
     const Device& device;
     Scheduler& scheduler;
     DescriptorPool& descriptor_pool;
-    UpdateDescriptorQueue& update_descriptor_queue;
+    GuestDescriptorQueue& guest_descriptor_queue;
     RenderPassCache& render_pass_cache;
     BufferCache& buffer_cache;
     TextureCache& texture_cache;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -160,17 +160,16 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
     : RasterizerAccelerated{cpu_memory_}, gpu{gpu_}, screen_info{screen_info_}, device{device_},
       memory_allocator{memory_allocator_}, state_tracker{state_tracker_}, scheduler{scheduler_},
       staging_pool(device, memory_allocator, scheduler), descriptor_pool(device, scheduler),
-      update_descriptor_queue(device, scheduler),
-      blit_image(device, scheduler, state_tracker, descriptor_pool),
-      render_pass_cache(device), texture_cache_runtime{device,           scheduler,
-                                                       memory_allocator, staging_pool,
-                                                       blit_image,       render_pass_cache,
-                                                       descriptor_pool,  update_descriptor_queue},
+      guest_descriptor_queue(device, scheduler), compute_pass_descriptor_queue(device, scheduler),
+      blit_image(device, scheduler, state_tracker, descriptor_pool), render_pass_cache(device),
+      texture_cache_runtime{
+          device,     scheduler,         memory_allocator, staging_pool,
+          blit_image, render_pass_cache, descriptor_pool,  compute_pass_descriptor_queue},
       texture_cache(texture_cache_runtime, *this),
       buffer_cache_runtime(device, memory_allocator, scheduler, staging_pool,
-                           update_descriptor_queue, descriptor_pool),
+                           guest_descriptor_queue, compute_pass_descriptor_queue, descriptor_pool),
       buffer_cache(*this, cpu_memory_, buffer_cache_runtime),
-      pipeline_cache(*this, device, scheduler, descriptor_pool, update_descriptor_queue,
+      pipeline_cache(*this, device, scheduler, descriptor_pool, guest_descriptor_queue,
                      render_pass_cache, buffer_cache, texture_cache, gpu.ShaderNotify()),
       query_cache{*this, cpu_memory_, device, scheduler},
       accelerate_dma(buffer_cache, texture_cache, scheduler),
@@ -349,25 +348,12 @@ void RasterizerVulkan::Clear(u32 layer_count) {
 
     const u32 color_attachment = regs.clear_surface.RT;
     if (use_color && framebuffer->HasAspectColorBit(color_attachment)) {
-        VkClearValue clear_value;
-        bool is_integer = false;
-        bool is_signed = false;
-        size_t int_size = 8;
-        for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; ++i) {
-            const auto& this_rt = regs.rt[i];
-            if (this_rt.Address() == 0) {
-                continue;
-            }
-            if (this_rt.format == Tegra::RenderTargetFormat::NONE) {
-                continue;
-            }
-            const auto format =
-                VideoCore::Surface::PixelFormatFromRenderTargetFormat(this_rt.format);
-            is_integer = IsPixelFormatInteger(format);
-            is_signed = IsPixelFormatSignedInteger(format);
-            int_size = PixelComponentSizeBitsInteger(format);
-            break;
-        }
+        const auto format =
+            VideoCore::Surface::PixelFormatFromRenderTargetFormat(regs.rt[color_attachment].format);
+        bool is_integer = IsPixelFormatInteger(format);
+        bool is_signed = IsPixelFormatSignedInteger(format);
+        size_t int_size = PixelComponentSizeBitsInteger(format);
+        VkClearValue clear_value{};
         if (!is_integer) {
             std::memcpy(clear_value.color.float32, regs.clear_color.data(),
                         regs.clear_color.size() * sizeof(f32));
@@ -502,6 +488,22 @@ bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
+VideoCore::RasterizerDownloadArea RasterizerVulkan::GetFlushArea(VAddr addr, u64 size) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        auto area = texture_cache.GetFlushArea(addr, size);
+        if (area) {
+            return *area;
+        }
+    }
+    VideoCore::RasterizerDownloadArea new_area{
+        .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
+        .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
+        .preemtive = true,
+    };
+    return new_area;
+}
+
 void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
@@ -598,7 +600,7 @@ void RasterizerVulkan::SignalSyncPoint(u32 value) {
 }
 
 void RasterizerVulkan::SignalReference() {
-    fence_manager.SignalOrdering();
+    fence_manager.SignalReference();
 }
 
 void RasterizerVulkan::ReleaseFences() {
@@ -631,7 +633,7 @@ void RasterizerVulkan::WaitForIdle() {
         cmdbuf.SetEvent(event, flags);
         cmdbuf.WaitEvents(event, flags, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, {}, {}, {});
     });
-    SignalReference();
+    fence_manager.SignalOrdering();
 }
 
 void RasterizerVulkan::FragmentBarrier() {
@@ -653,7 +655,8 @@ void RasterizerVulkan::FlushCommands() {
 
 void RasterizerVulkan::TickFrame() {
     draw_counter = 0;
-    update_descriptor_queue.TickFrame();
+    guest_descriptor_queue.TickFrame();
+    compute_pass_descriptor_queue.TickFrame();
     fence_manager.TickFrame();
     staging_pool.TickFrame();
     {
@@ -777,7 +780,7 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
                                        const Tegra::DMA::BufferOperand& buffer_operand,
                                        const Tegra::DMA::ImageOperand& image_operand) {
     std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
-    const auto image_id = texture_cache.DmaImageId(image_operand);
+    const auto image_id = texture_cache.DmaImageId(image_operand, IS_IMAGE_UPLOAD);
     if (image_id == VideoCommon::NULL_IMAGE_ID) {
         return false;
     }

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -92,6 +92,7 @@ public:
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     bool MustFlushRegion(VAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) override;
@@ -183,7 +184,8 @@ private:
 
     StagingBufferPool staging_pool;
     DescriptorPool descriptor_pool;
-    UpdateDescriptorQueue update_descriptor_queue;
+    GuestDescriptorQueue guest_descriptor_queue;
+    ComputePassDescriptorQueue compute_pass_descriptor_queue;
     BlitImageHelper blit_image;
     RenderPassCache render_pass_cache;
 

src/video_core/renderer_vulkan/vk_swapchain.cpp

@@ -34,8 +34,8 @@ VkSurfaceFormatKHR ChooseSwapSurfaceFormat(vk::Span<VkSurfaceFormatKHR> formats)
     return found != formats.end() ? *found : formats[0];
 }
 
-static constexpr VkPresentModeKHR ChooseSwapPresentMode(bool has_imm, bool has_mailbox,
-                                                        bool has_fifo_relaxed) {
+static VkPresentModeKHR ChooseSwapPresentMode(bool has_imm, bool has_mailbox,
+                                              bool has_fifo_relaxed) {
     // Mailbox doesn't lock the application like FIFO (vsync)
     // FIFO present mode locks the framerate to the monitor's refresh rate
     Settings::VSyncMode setting = [has_imm, has_mailbox]() {