renderer_vulkan: Remove timeline semaphore wait

video_core: tune garbage collection aggressiveness

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/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

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

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};
@@ -472,6 +478,10 @@ struct Values {
     SwitchableSetting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
     SwitchableSetting<bool> use_vulkan_driver_pipeline_cache{true,
                                                              "use_vulkan_driver_pipeline_cache"};
+    SwitchableSetting<bool> enable_compute_pipelines{false, "enable_compute_pipelines"};
+    SwitchableSetting<AstcRecompression, true> astc_recompression{
+        AstcRecompression::Uncompressed, AstcRecompression::Uncompressed, AstcRecompression::Bc3,
+        "astc_recompression"};
 
     SwitchableSetting<u8> bg_red{0, "bg_red"};
     SwitchableSetting<u8> bg_green{0, "bg_green"};

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/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

@@ -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/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/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) {
@@ -398,8 +398,9 @@ 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_protocol/common_protocol.cpp

@@ -265,7 +265,7 @@ DriverResult JoyconCommonProtocol::SendMCUData(ReportMode report_mode, MCUSubCom
 
 DriverResult JoyconCommonProtocol::WaitSetMCUMode(ReportMode report_mode, MCUMode mode) {
     MCUCommandResponse output{};
-    constexpr std::size_t MaxTries{8};
+    constexpr std::size_t MaxTries{16};
     std::size_t tries{};
 
     do {

src/input_common/helpers/joycon_protocol/joycon_types.h

@@ -394,6 +394,7 @@ enum class DriverResult {
     InvalidHandle,
     NotSupported,
     Disabled,
+    Delayed,
     Unknown,
 };
 
@@ -576,8 +577,8 @@ static_assert(sizeof(NFCPollingCommandData) == 0x05, "NFCPollingCommandData is a
 
 struct NFCRequestState {
     NFCReadCommand command_argument;
-    u8 packet_id;
     INSERT_PADDING_BYTES(0x1);
+    u8 packet_id;
     MCUPacketFlag packet_flag;
     u8 data_length;
     union {

src/input_common/helpers/joycon_protocol/nfc.cpp

@@ -64,6 +64,20 @@ DriverResult NfcProtocol::StartNFCPollingMode() {
     if (result == DriverResult::Success) {
         result = WaitUntilNfcIsReady();
     }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStopPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIsReady();
+    }
+    if (result == DriverResult::Success) {
+        MCUCommandResponse output{};
+        result = SendStartPollingRequest(output);
+    }
+    if (result == DriverResult::Success) {
+        result = WaitUntilNfcIsPolling();
+    }
     if (result == DriverResult::Success) {
         is_enabled = true;
     }
@@ -72,24 +86,26 @@ 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);
     }
 
@@ -97,12 +113,17 @@ DriverResult NfcProtocol::ScanAmiibo(std::vector<u8>& data) {
 }
 
 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;
@@ -114,7 +135,7 @@ DriverResult NfcProtocol::WaitUntilNfcIsReady() {
     std::size_t tries = 0;
 
     do {
-        auto result = SendStartWaitingRecieveRequest(output);
+        auto result = SendNextPackageRequest(output, {});
 
         if (result != DriverResult::Success) {
             return result;
@@ -129,14 +150,14 @@ DriverResult NfcProtocol::WaitUntilNfcIsReady() {
     return DriverResult::Success;
 }
 
-DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
-    LOG_DEBUG(Input, "Start Polling for tag");
-    constexpr std::size_t timeout_limit = 7;
+DriverResult NfcProtocol::WaitUntilNfcIsPolling() {
+    constexpr std::size_t timeout_limit = 10;
     MCUCommandResponse output{};
     std::size_t tries = 0;
 
     do {
-        const auto result = SendStartPollingRequest(output);
+        auto result = SendNextPackageRequest(output, {});
+
         if (result != DriverResult::Success) {
             return result;
         }
@@ -145,7 +166,26 @@ 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[5] != 0x31 || output.mcu_data[6] != 0x01);
+
+    return DriverResult::Success;
+}
+
+DriverResult NfcProtocol::IsTagInRange(TagFoundData& data, std::size_t timeout_limit) {
+    MCUCommandResponse output{};
+    std::size_t tries = 0;
+
+    do {
+        const auto result = SendNextPackageRequest(output, {});
+        if (result != DriverResult::Success) {
+            return result;
+        }
+        if (tries++ > timeout_limit) {
+            return DriverResult::Timeout;
+        }
+    } 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] != 0x04));
 
     data.type = output.mcu_data[12];
     data.uuid.resize(output.mcu_data[14]);
@@ -154,85 +194,22 @@ DriverResult NfcProtocol::StartPolling(TagFoundData& data) {
     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 +232,7 @@ 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;
         }
 
@@ -262,18 +240,9 @@ 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) {
-            continue;
-        }
-
-        if (tries++ > timeout_limit) {
-            return DriverResult::Timeout;
-        }
     }
 
-    return DriverResult::Success;
+    return DriverResult::Timeout;
 }
 
 DriverResult NfcProtocol::SendStartPollingRequest(MCUCommandResponse& output) {
@@ -317,10 +286,10 @@ DriverResult NfcProtocol::SendStopPollingRequest(MCUCommandResponse& output) {
                        output);
 }
 
-DriverResult NfcProtocol::SendStartWaitingRecieveRequest(MCUCommandResponse& output) {
+DriverResult NfcProtocol::SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id) {
     NFCRequestState request{
         .command_argument = NFCReadCommand::StartWaitingRecieve,
-        .packet_id = 0x0,
+        .packet_id = packet_id,
         .packet_flag = MCUPacketFlag::LastCommandPacket,
         .data_length = 0,
         .raw_data = {},

src/input_common/helpers/joycon_protocol/nfc.h

@@ -32,6 +32,9 @@ public:
     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;
@@ -39,9 +42,9 @@ private:
 
     DriverResult WaitUntilNfcIsReady();
 
-    DriverResult StartPolling(TagFoundData& data);
+    DriverResult WaitUntilNfcIsPolling();
 
-    DriverResult ReadTag(const TagFoundData& data);
+    DriverResult IsTagInRange(TagFoundData& data, std::size_t timeout_limit = 1);
 
     DriverResult GetAmiiboData(std::vector<u8>& data);
 
@@ -49,13 +52,14 @@ private:
 
     DriverResult SendStopPollingRequest(MCUCommandResponse& output);
 
-    DriverResult SendStartWaitingRecieveRequest(MCUCommandResponse& output);
+    DriverResult SendNextPackageRequest(MCUCommandResponse& output, u8 packet_id);
 
     DriverResult SendReadAmiiboRequest(MCUCommandResponse& output, NFCPages ntag_pages);
 
     NFCReadBlockCommand GetReadBlockCommand(NFCPages pages) 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/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/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_cache.h

@@ -30,8 +30,8 @@ BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
     }
 
     const s64 device_memory = static_cast<s64>(runtime.GetDeviceLocalMemory());
-    const s64 min_spacing_expected = device_memory - 1_GiB - 512_MiB;
-    const s64 min_spacing_critical = device_memory - 1_GiB;
+    const s64 min_spacing_expected = device_memory - 1_GiB;
+    const s64 min_spacing_critical = device_memory - 512_MiB;
     const s64 mem_threshold = std::min(device_memory, TARGET_THRESHOLD);
     const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
     const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
@@ -96,12 +96,12 @@ void BufferCache<P>::TickFrame() {
 
 template <class P>
 void BufferCache<P>::WriteMemory(VAddr cpu_addr, u64 size) {
-    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
     if (memory_tracker.IsRegionGpuModified(cpu_addr, size)) {
         const IntervalType subtract_interval{cpu_addr, cpu_addr + size};
         ClearDownload(subtract_interval);
         common_ranges.subtract(subtract_interval);
     }
+    memory_tracker.MarkRegionAsCpuModified(cpu_addr, size);
 }
 
 template <class P>
@@ -122,41 +122,24 @@ std::optional<VideoCore::RasterizerDownloadArea> BufferCache<P>::GetFlushArea(VA
         area->preemtive = true;
         return area;
     };
+    area->preemtive =
+        !IsRegionGpuModified(cpu_addr_start_aligned, cpu_addr_end_aligned - cpu_addr_start_aligned);
     memory_tracker.MarkRegionAsPreflushable(cpu_addr_start_aligned,
                                             cpu_addr_end_aligned - cpu_addr_start_aligned);
-    area->preemtive = !IsRegionGpuModified(cpu_addr, size);
     return area;
 }
 
 template <class P>
 void BufferCache<P>::DownloadMemory(VAddr cpu_addr, u64 size) {
-    WaitOnAsyncFlushes(cpu_addr, size);
     ForEachBufferInRange(cpu_addr, size, [&](BufferId, Buffer& buffer) {
         DownloadBufferMemory(buffer, cpu_addr, size);
     });
 }
 
-template <class P>
-void BufferCache<P>::WaitOnAsyncFlushes(VAddr cpu_addr, u64 size) {
-    bool must_wait = false;
-    ForEachInOverlapCounter(async_downloads, cpu_addr, size,
-                            [&](VAddr, VAddr, int) { must_wait = true; });
-    bool must_release = false;
-    ForEachInRangeSet(pending_ranges, cpu_addr, size, [&](VAddr, VAddr) { must_release = true; });
-    if (must_release) {
-        std::function<void()> tmp([]() {});
-        rasterizer.SignalFence(std::move(tmp));
-    }
-    if (must_wait || must_release) {
-        rasterizer.ReleaseFences();
-    }
-}
-
 template <class P>
 void BufferCache<P>::ClearDownload(IntervalType subtract_interval) {
     RemoveEachInOverlapCounter(async_downloads, subtract_interval, -1024);
     uncommitted_ranges.subtract(subtract_interval);
-    pending_ranges.subtract(subtract_interval);
     for (auto& interval_set : committed_ranges) {
         interval_set.subtract(subtract_interval);
     }
@@ -176,7 +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;
@@ -204,7 +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.
@@ -491,7 +472,6 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
     }
     MICROPROFILE_SCOPE(GPU_DownloadMemory);
 
-    pending_ranges.clear();
     auto it = committed_ranges.begin();
     while (it != committed_ranges.end()) {
         auto& current_intervals = *it;
@@ -1223,16 +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>
@@ -1677,15 +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

@@ -381,8 +381,6 @@ private:
 
     void RunGarbageCollector();
 
-    void WaitOnAsyncFlushes(VAddr cpu_addr, u64 size);
-
     void BindHostIndexBuffer();
 
     void BindHostVertexBuffers();
@@ -547,7 +545,6 @@ private:
     IntervalSet uncommitted_ranges;
     IntervalSet common_ranges;
     IntervalSet cached_ranges;
-    IntervalSet pending_ranges;
     std::deque<IntervalSet> committed_ranges;
 
     // Async Buffers

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/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/renderer_opengl/gl_texture_cache.cpp

@@ -233,6 +233,8 @@ void ApplySwizzle(GLuint handle, PixelFormat format, std::array<SwizzleSource, 4
                                     const VideoCommon::ImageInfo& info) {
     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;
     }
 
@@ -123,15 +128,12 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
     const std::array signal_values{host_tick, u64(0)};
     const std::array signal_semaphores{timeline_semaphore, signal_semaphore};
 
-    const u32 num_wait_semaphores = wait_semaphore ? 2 : 1;
-    const std::array wait_values{host_tick - 1, u64(1)};
-    const std::array wait_semaphores{timeline_semaphore, wait_semaphore};
-
+    const u32 num_wait_semaphores = wait_semaphore ? 1 : 0;
     const VkTimelineSemaphoreSubmitInfo timeline_si{
         .sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO,
         .pNext = nullptr,
-        .waitSemaphoreValueCount = num_wait_semaphores,
-        .pWaitSemaphoreValues = wait_values.data(),
+        .waitSemaphoreValueCount = 0,
+        .pWaitSemaphoreValues = nullptr,
         .signalSemaphoreValueCount = num_signal_semaphores,
         .pSignalSemaphoreValues = signal_values.data(),
     };
@@ -139,7 +141,7 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
         .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
         .pNext = &timeline_si,
         .waitSemaphoreCount = num_wait_semaphores,
-        .pWaitSemaphores = wait_semaphores.data(),
+        .pWaitSemaphores = &wait_semaphore,
         .pWaitDstStageMask = wait_stage_masks.data(),
         .commandBufferCount = 1,
         .pCommandBuffers = cmdbuf.address(),
@@ -167,16 +169,53 @@ 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));
@@ -669,7 +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();
     {

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -184,7 +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]() {

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -798,13 +798,13 @@ TextureCacheRuntime::TextureCacheRuntime(const Device& device_, Scheduler& sched
                                          BlitImageHelper& blit_image_helper_,
                                          RenderPassCache& render_pass_cache_,
                                          DescriptorPool& descriptor_pool,
-                                         UpdateDescriptorQueue& update_descriptor_queue)
+                                         ComputePassDescriptorQueue& compute_pass_descriptor_queue)
     : device{device_}, scheduler{scheduler_}, memory_allocator{memory_allocator_},
       staging_buffer_pool{staging_buffer_pool_}, blit_image_helper{blit_image_helper_},
       render_pass_cache{render_pass_cache_}, resolution{Settings::values.resolution_info} {
     if (Settings::values.accelerate_astc) {
         astc_decoder_pass.emplace(device, scheduler, descriptor_pool, staging_buffer_pool,
-                                  update_descriptor_queue, memory_allocator);
+                                  compute_pass_descriptor_queue, memory_allocator);
     }
 }
 
@@ -861,6 +861,10 @@ VkBuffer TextureCacheRuntime::GetTemporaryBuffer(size_t needed_size) {
     return *buffers[level];
 }
 
+void TextureCacheRuntime::BarrierFeedbackLoop() {
+    scheduler.RequestOutsideRenderPassOperationContext();
+}
+
 void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
                                            std::span<const VideoCommon::ImageCopy> copies) {
     std::vector<VkBufferImageCopy> vk_in_copies(copies.size());
@@ -1268,7 +1272,9 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
     if (IsPixelFormatASTC(info.format) && !runtime->device.IsOptimalAstcSupported()) {
         if (Settings::values.async_astc.GetValue()) {
             flags |= VideoCommon::ImageFlagBits::AsynchronousDecode;
-        } else if (Settings::values.accelerate_astc.GetValue() && info.size.depth == 1) {
+        } else if (Settings::values.astc_recompression.GetValue() ==
+                       Settings::AstcRecompression::Uncompressed &&
+                   Settings::values.accelerate_astc.GetValue() && info.size.depth == 1) {
             flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
         }
         flags |= VideoCommon::ImageFlagBits::Converted;
@@ -1283,7 +1289,9 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
         .usage = VK_IMAGE_USAGE_STORAGE_BIT,
     };
     current_image = *original_image;
-    if (IsPixelFormatASTC(info.format) && !runtime->device.IsOptimalAstcSupported()) {
+    if (IsPixelFormatASTC(info.format) && !runtime->device.IsOptimalAstcSupported() &&
+        Settings::values.astc_recompression.GetValue() ==
+            Settings::AstcRecompression::Uncompressed) {
         const auto& device = runtime->device.GetLogical();
         storage_image_views.reserve(info.resources.levels);
         for (s32 level = 0; level < info.resources.levels; ++level) {
@@ -1864,6 +1872,7 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
         num_layers = std::max(num_layers, color_buffer->range.extent.layers);
         images[num_images] = color_buffer->ImageHandle();
         image_ranges[num_images] = MakeSubresourceRange(color_buffer);
+        rt_map[index] = num_images;
         samples = color_buffer->Samples();
         ++num_images;
     }

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -34,7 +34,6 @@ class ImageView;
 class Framebuffer;
 class RenderPassCache;
 class StagingBufferPool;
-class UpdateDescriptorQueue;
 class Scheduler;
 
 class TextureCacheRuntime {
@@ -45,7 +44,7 @@ public:
                                  BlitImageHelper& blit_image_helper_,
                                  RenderPassCache& render_pass_cache_,
                                  DescriptorPool& descriptor_pool,
-                                 UpdateDescriptorQueue& update_descriptor_queue);
+                                 ComputePassDescriptorQueue& compute_pass_descriptor_queue);
 
     void Finish();
 
@@ -104,6 +103,8 @@ public:
 
     [[nodiscard]] VkBuffer GetTemporaryBuffer(size_t needed_size);
 
+    void BarrierFeedbackLoop();
+
     const Device& device;
     Scheduler& scheduler;
     MemoryAllocator& memory_allocator;
@@ -335,7 +336,7 @@ public:
     }
 
     [[nodiscard]] bool HasAspectColorBit(size_t index) const noexcept {
-        return (image_ranges.at(index).aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) != 0;
+        return (image_ranges.at(rt_map[index]).aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) != 0;
     }
 
     [[nodiscard]] bool HasAspectDepthBit() const noexcept {
@@ -355,6 +356,7 @@ private:
     u32 num_images = 0;
     std::array<VkImage, 9> images{};
     std::array<VkImageSubresourceRange, 9> image_ranges{};
+    std::array<size_t, NUM_RT> rt_map{};
     bool has_depth{};
     bool has_stencil{};
 };

src/video_core/renderer_vulkan/vk_update_descriptor.h

@@ -32,7 +32,7 @@ class UpdateDescriptorQueue final {
     // This should be plenty for the vast majority of cases. Most desktop platforms only
     // provide up to 3 swapchain images.
     static constexpr size_t FRAMES_IN_FLIGHT = 5;
-    static constexpr size_t FRAME_PAYLOAD_SIZE = 0x10000;
+    static constexpr size_t FRAME_PAYLOAD_SIZE = 0x20000;
     static constexpr size_t PAYLOAD_SIZE = FRAME_PAYLOAD_SIZE * FRAMES_IN_FLIGHT;
 
 public:
@@ -86,4 +86,8 @@ private:
     std::array<DescriptorUpdateEntry, PAYLOAD_SIZE> payload;
 };
 
+// TODO: should these be separate classes instead?
+using GuestDescriptorQueue = UpdateDescriptorQueue;
+using ComputePassDescriptorQueue = UpdateDescriptorQueue;
+
 } // namespace Vulkan

src/video_core/texture_cache/texture_cache.h

@@ -49,8 +49,8 @@ TextureCache<P>::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface&
 
     if constexpr (HAS_DEVICE_MEMORY_INFO) {
         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;
@@ -86,10 +86,12 @@ void TextureCache<P>::RunGarbageCollector() {
             // used by the async decoder thread.
             return false;
         }
+        if (!aggressive_mode && True(image.flags & ImageFlagBits::CostlyLoad)) {
+            return false;
+        }
         const bool must_download =
             image.IsSafeDownload() && False(image.flags & ImageFlagBits::BadOverlap);
-        if (!high_priority_mode &&
-            (must_download || True(image.flags & ImageFlagBits::CostlyLoad))) {
+        if (!high_priority_mode && must_download) {
             return false;
         }
         if (must_download) {
@@ -183,6 +185,42 @@ void TextureCache<P>::FillComputeImageViews(std::span<ImageViewInOut> views) {
                          views);
 }
 
+template <class P>
+void TextureCache<P>::CheckFeedbackLoop(std::span<const ImageViewInOut> views) {
+    const bool requires_barrier = [&] {
+        for (const auto& view : views) {
+            if (!view.id) {
+                continue;
+            }
+            auto& image_view = slot_image_views[view.id];
+
+            // Check color targets
+            for (const auto& ct_view_id : render_targets.color_buffer_ids) {
+                if (ct_view_id) {
+                    auto& ct_view = slot_image_views[ct_view_id];
+                    if (image_view.image_id == ct_view.image_id) {
+                        return true;
+                    }
+                }
+            }
+
+            // Check zeta target
+            if (render_targets.depth_buffer_id) {
+                auto& zt_view = slot_image_views[render_targets.depth_buffer_id];
+                if (image_view.image_id == zt_view.image_id) {
+                    return true;
+                }
+            }
+        }
+
+        return false;
+    }();
+
+    if (requires_barrier) {
+        runtime.BarrierFeedbackLoop();
+    }
+}
+
 template <class P>
 typename P::Sampler* TextureCache<P>::GetGraphicsSampler(u32 index) {
     if (index > channel_state->graphics_sampler_table.Limit()) {
@@ -1469,7 +1507,7 @@ std::optional<typename TextureCache<P>::BlitImages> TextureCache<P>::GetBlitImag
         if (!copy.must_accelerate) {
             do {
                 if (!src_id && !dst_id) {
-                    return std::nullopt;
+                    break;
                 }
                 if (src_id && True(slot_images[src_id].flags & ImageFlagBits::GpuModified)) {
                     break;

src/video_core/texture_cache/texture_cache_base.h

@@ -148,6 +148,9 @@ public:
     /// Fill image_view_ids with the compute images in indices
     void FillComputeImageViews(std::span<ImageViewInOut> views);
 
+    /// Handle feedback loops during draws.
+    void CheckFeedbackLoop(std::span<const ImageViewInOut> views);
+
     /// Get the sampler from the graphics descriptor table in the specified index
     Sampler* GetGraphicsSampler(u32 index);
 

src/video_core/texture_cache/util.cpp

@@ -18,6 +18,8 @@
 #include "common/bit_util.h"
 #include "common/common_types.h"
 #include "common/div_ceil.h"
+#include "common/scratch_buffer.h"
+#include "common/settings.h"
 #include "video_core/compatible_formats.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
@@ -28,6 +30,7 @@
 #include "video_core/texture_cache/samples_helper.h"
 #include "video_core/texture_cache/util.h"
 #include "video_core/textures/astc.h"
+#include "video_core/textures/bcn.h"
 #include "video_core/textures/decoders.h"
 
 namespace VideoCommon {
@@ -585,6 +588,21 @@ u32 CalculateConvertedSizeBytes(const ImageInfo& info) noexcept {
         return info.size.width * BytesPerBlock(info.format);
     }
     static constexpr Extent2D TILE_SIZE{1, 1};
+    if (IsPixelFormatASTC(info.format) && Settings::values.astc_recompression.GetValue() !=
+                                              Settings::AstcRecompression::Uncompressed) {
+        const u32 bpp_div =
+            Settings::values.astc_recompression.GetValue() == Settings::AstcRecompression::Bc1 ? 2
+                                                                                               : 1;
+        // NumBlocksPerLayer doesn't account for this correctly, so we have to do it manually.
+        u32 output_size = 0;
+        for (s32 i = 0; i < info.resources.levels; i++) {
+            const auto mip_size = AdjustMipSize(info.size, i);
+            const u32 plane_dim =
+                Common::AlignUp(mip_size.width, 4U) * Common::AlignUp(mip_size.height, 4U);
+            output_size += (plane_dim * info.size.depth * info.resources.layers) / bpp_div;
+        }
+        return output_size;
+    }
     return NumBlocksPerLayer(info, TILE_SIZE) * info.resources.layers * CONVERTED_BYTES_PER_BLOCK;
 }
 
@@ -885,6 +903,7 @@ BufferCopy UploadBufferCopy(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr,
 void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8> output,
                   std::span<BufferImageCopy> copies) {
     u32 output_offset = 0;
+    Common::ScratchBuffer<u8> decode_scratch;
 
     const Extent2D tile_size = DefaultBlockSize(info.format);
     for (BufferImageCopy& copy : copies) {
@@ -895,22 +914,58 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
         ASSERT(copy.image_extent == mip_size);
         ASSERT(copy.buffer_row_length == Common::AlignUp(mip_size.width, tile_size.width));
         ASSERT(copy.buffer_image_height == Common::AlignUp(mip_size.height, tile_size.height));
-        if (IsPixelFormatASTC(info.format)) {
-            Tegra::Texture::ASTC::Decompress(
-                input.subspan(copy.buffer_offset), copy.image_extent.width,
-                copy.image_extent.height,
-                copy.image_subresource.num_layers * copy.image_extent.depth, tile_size.width,
-                tile_size.height, output.subspan(output_offset));
-        } else {
-            DecompressBC4(input.subspan(copy.buffer_offset), copy.image_extent,
-                          output.subspan(output_offset));
-        }
+
+        const auto input_offset = input.subspan(copy.buffer_offset);
         copy.buffer_offset = output_offset;
         copy.buffer_row_length = mip_size.width;
         copy.buffer_image_height = mip_size.height;
 
-        output_offset += copy.image_extent.width * copy.image_extent.height *
-                         copy.image_subresource.num_layers * CONVERTED_BYTES_PER_BLOCK;
+        const auto recompression_setting = Settings::values.astc_recompression.GetValue();
+        const bool astc = IsPixelFormatASTC(info.format);
+
+        if (astc && recompression_setting == Settings::AstcRecompression::Uncompressed) {
+            Tegra::Texture::ASTC::Decompress(
+                input_offset, copy.image_extent.width, copy.image_extent.height,
+                copy.image_subresource.num_layers * copy.image_extent.depth, tile_size.width,
+                tile_size.height, output.subspan(output_offset));
+
+            output_offset += copy.image_extent.width * copy.image_extent.height *
+                             copy.image_subresource.num_layers * CONVERTED_BYTES_PER_BLOCK;
+        } else if (astc) {
+            // BC1 uses 0.5 bytes per texel
+            // BC3 uses 1 byte per texel
+            const auto compress = recompression_setting == Settings::AstcRecompression::Bc1
+                                      ? Tegra::Texture::BCN::CompressBC1
+                                      : Tegra::Texture::BCN::CompressBC3;
+            const auto bpp_div = recompression_setting == Settings::AstcRecompression::Bc1 ? 2 : 1;
+
+            const u32 plane_dim = copy.image_extent.width * copy.image_extent.height;
+            const u32 level_size = plane_dim * copy.image_extent.depth *
+                                   copy.image_subresource.num_layers * CONVERTED_BYTES_PER_BLOCK;
+            decode_scratch.resize_destructive(level_size);
+
+            Tegra::Texture::ASTC::Decompress(
+                input_offset, copy.image_extent.width, copy.image_extent.height,
+                copy.image_subresource.num_layers * copy.image_extent.depth, tile_size.width,
+                tile_size.height, decode_scratch);
+
+            compress(decode_scratch, copy.image_extent.width, copy.image_extent.height,
+                     copy.image_subresource.num_layers * copy.image_extent.depth,
+                     output.subspan(output_offset));
+
+            const u32 aligned_plane_dim = Common::AlignUp(copy.image_extent.width, 4) *
+                                          Common::AlignUp(copy.image_extent.height, 4);
+
+            copy.buffer_size =
+                (aligned_plane_dim * copy.image_extent.depth * copy.image_subresource.num_layers) /
+                bpp_div;
+            output_offset += static_cast<u32>(copy.buffer_size);
+        } else {
+            DecompressBC4(input_offset, copy.image_extent, output.subspan(output_offset));
+
+            output_offset += copy.image_extent.width * copy.image_extent.height *
+                             copy.image_subresource.num_layers * CONVERTED_BYTES_PER_BLOCK;
+        }
     }
 }
 

src/video_core/textures/astc.cpp

@@ -16,8 +16,8 @@
 #include "common/alignment.h"
 #include "common/common_types.h"
 #include "common/polyfill_ranges.h"
-#include "common/thread_worker.h"
 #include "video_core/textures/astc.h"
+#include "video_core/textures/workers.h"
 
 class InputBitStream {
 public:
@@ -1656,8 +1656,7 @@ void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height,
     const u32 rows = Common::DivideUp(height, block_height);
     const u32 cols = Common::DivideUp(width, block_width);
 
-    static Common::ThreadWorker workers{std::max(std::thread::hardware_concurrency(), 2U) / 2,
-                                        "ASTCDecompress"};
+    Common::ThreadWorker& workers{GetThreadWorkers()};
 
     for (u32 z = 0; z < depth; ++z) {
         const u32 depth_offset = z * height * width * 4;

src/video_core/textures/bcn.cpp

@@ -0,0 +1,87 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <stb_dxt.h>
+#include <string.h>
+
+#include "common/alignment.h"
+#include "video_core/textures/bcn.h"
+#include "video_core/textures/workers.h"
+
+namespace Tegra::Texture::BCN {
+
+using BCNCompressor = void(u8* block_output, const u8* block_input, bool any_alpha);
+
+template <u32 BytesPerBlock, bool ThresholdAlpha = false>
+void CompressBCN(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
+                 std::span<uint8_t> output, BCNCompressor f) {
+    constexpr u8 alpha_threshold = 128;
+    constexpr u32 bytes_per_px = 4;
+    const u32 plane_dim = width * height;
+
+    Common::ThreadWorker& workers{GetThreadWorkers()};
+
+    for (u32 z = 0; z < depth; z++) {
+        for (u32 y = 0; y < height; y += 4) {
+            auto compress_row = [z, y, width, height, plane_dim, f, data, output]() {
+                for (u32 x = 0; x < width; x += 4) {
+                    // Gather 4x4 block of RGBA texels
+                    u8 input_colors[4][4][4];
+                    bool any_alpha = false;
+
+                    for (u32 j = 0; j < 4; j++) {
+                        for (u32 i = 0; i < 4; i++) {
+                            const size_t coord =
+                                (z * plane_dim + (y + j) * width + (x + i)) * bytes_per_px;
+
+                            if ((x + i < width) && (y + j < height)) {
+                                if constexpr (ThresholdAlpha) {
+                                    if (data[coord + 3] >= alpha_threshold) {
+                                        input_colors[j][i][0] = data[coord + 0];
+                                        input_colors[j][i][1] = data[coord + 1];
+                                        input_colors[j][i][2] = data[coord + 2];
+                                        input_colors[j][i][3] = 255;
+                                    } else {
+                                        any_alpha = true;
+                                        memset(input_colors[j][i], 0, bytes_per_px);
+                                    }
+                                } else {
+                                    memcpy(input_colors[j][i], &data[coord], bytes_per_px);
+                                }
+                            } else {
+                                memset(input_colors[j][i], 0, bytes_per_px);
+                            }
+                        }
+                    }
+
+                    const u32 bytes_per_row = BytesPerBlock * Common::DivideUp(width, 4U);
+                    const u32 bytes_per_plane = bytes_per_row * Common::DivideUp(height, 4U);
+                    f(output.data() + z * bytes_per_plane + (y / 4) * bytes_per_row +
+                          (x / 4) * BytesPerBlock,
+                      reinterpret_cast<u8*>(input_colors), any_alpha);
+                }
+            };
+            workers.QueueWork(std::move(compress_row));
+        }
+        workers.WaitForRequests();
+    }
+}
+
+void CompressBC1(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
+                 std::span<uint8_t> output) {
+    CompressBCN<8, true>(data, width, height, depth, output,
+                         [](u8* block_output, const u8* block_input, bool any_alpha) {
+                             stb_compress_bc1_block(block_output, block_input, any_alpha,
+                                                    STB_DXT_NORMAL);
+                         });
+}
+
+void CompressBC3(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
+                 std::span<uint8_t> output) {
+    CompressBCN<16, false>(data, width, height, depth, output,
+                           [](u8* block_output, const u8* block_input, bool any_alpha) {
+                               stb_compress_bc3_block(block_output, block_input, STB_DXT_NORMAL);
+                           });
+}
+
+} // namespace Tegra::Texture::BCN

src/video_core/textures/bcn.h

@@ -0,0 +1,17 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <span>
+#include <stdint.h>
+
+namespace Tegra::Texture::BCN {
+
+void CompressBC1(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
+                 std::span<uint8_t> output);
+
+void CompressBC3(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
+                 std::span<uint8_t> output);
+
+} // namespace Tegra::Texture::BCN

src/video_core/textures/workers.cpp

@@ -0,0 +1,15 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "video_core/textures/workers.h"
+
+namespace Tegra::Texture {
+
+Common::ThreadWorker& GetThreadWorkers() {
+    static Common::ThreadWorker workers{std::max(std::thread::hardware_concurrency(), 2U) / 2,
+                                        "ImageTranscode"};
+
+    return workers;
+}
+
+} // namespace Tegra::Texture

src/video_core/textures/workers.h

@@ -0,0 +1,12 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/thread_worker.h"
+
+namespace Tegra::Texture {
+
+Common::ThreadWorker& GetThreadWorkers();
+
+}

src/video_core/vulkan_common/vulkan_device.cpp

@@ -406,6 +406,14 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable = false;
         features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation = false;
         dynamic_state3_blending = false;
+
+        const u32 version = (properties.properties.driverVersion << 3) >> 3;
+        if (version < VK_MAKE_API_VERSION(0, 23, 1, 0)) {
+            LOG_WARNING(Render_Vulkan,
+                        "RADV versions older than 23.1.0 have broken depth clamp dynamic state");
+            features.extended_dynamic_state3.extendedDynamicState3DepthClampEnable = false;
+            dynamic_state3_enables = false;
+        }
     }
     if (extensions.vertex_input_dynamic_state && is_radv) {
         // TODO(ameerj): Blacklist only offending driver versions
@@ -463,6 +471,17 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         LOG_WARNING(Render_Vulkan, "ANV driver does not support native BGR format");
         must_emulate_bgr565 = true;
     }
+    if (extensions.push_descriptor && is_intel_anv) {
+        const u32 version = (properties.properties.driverVersion << 3) >> 3;
+        if (version >= VK_MAKE_API_VERSION(0, 22, 3, 0)) {
+            // Disable VK_KHR_push_descriptor due to
+            // mesa/mesa/-/commit/ff91c5ca42bc80aa411cb3fd8f550aa6fdd16bdc
+            LOG_WARNING(Render_Vulkan,
+                        "ANV drivers 22.3.0 and later have broken VK_KHR_push_descriptor");
+            extensions.push_descriptor = false;
+            loaded_extensions.erase(VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
+        }
+    }
     if (is_mvk) {
         LOG_WARNING(Render_Vulkan,
                     "MVK driver breaks when using more than 16 vertex attributes/bindings");
@@ -982,6 +1001,11 @@ u64 Device::GetDeviceMemoryUsage() const {
 }
 
 void Device::CollectPhysicalMemoryInfo() {
+    // Account for resolution scaling in memory limits
+    const size_t normal_memory = 6_GiB;
+    const size_t scaler_memory = 1_GiB * Settings::values.resolution_info.ScaleUp(1);
+
+    // Calculate limits using memory budget
     VkPhysicalDeviceMemoryBudgetPropertiesEXT budget{};
     budget.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT;
     const auto mem_info =
@@ -1009,11 +1033,14 @@ void Device::CollectPhysicalMemoryInfo() {
         device_access_memory += mem_properties.memoryHeaps[element].size;
     }
     if (!is_integrated) {
+        const u64 reserve_memory = std::min<u64>(device_access_memory / 8, 1_GiB);
+        device_access_memory -= reserve_memory;
+        device_access_memory = std::min<u64>(device_access_memory, normal_memory + scaler_memory);
         return;
     }
     const s64 available_memory = static_cast<s64>(device_access_memory - device_initial_usage);
     device_access_memory = static_cast<u64>(std::max<s64>(
-        std::min<s64>(available_memory - 8_GiB, 4_GiB), static_cast<s64>(local_memory)));
+        std::min<s64>(available_memory - 8_GiB, 4_GiB), std::min<s64>(local_memory, 4_GiB)));
 }
 
 void Device::CollectToolingInfo() {

src/video_core/vulkan_common/vulkan_memory_allocator.cpp

@@ -147,7 +147,7 @@ public:
 
     /// Returns whether this allocation is compatible with the arguments.
     [[nodiscard]] bool IsCompatible(VkMemoryPropertyFlags flags, u32 type_mask) const {
-        return (flags & property_flags) == property_flags && (type_mask & shifted_memory_type) != 0;
+        return (flags & property_flags) == flags && (type_mask & shifted_memory_type) != 0;
     }
 
 private:

src/yuzu/configuration/config.cpp

@@ -711,11 +711,13 @@ void Config::ReadRendererValues() {
     ReadGlobalSetting(Settings::values.nvdec_emulation);
     ReadGlobalSetting(Settings::values.accelerate_astc);
     ReadGlobalSetting(Settings::values.async_astc);
+    ReadGlobalSetting(Settings::values.astc_recompression);
     ReadGlobalSetting(Settings::values.use_reactive_flushing);
     ReadGlobalSetting(Settings::values.shader_backend);
     ReadGlobalSetting(Settings::values.use_asynchronous_shaders);
     ReadGlobalSetting(Settings::values.use_fast_gpu_time);
     ReadGlobalSetting(Settings::values.use_vulkan_driver_pipeline_cache);
+    ReadGlobalSetting(Settings::values.enable_compute_pipelines);
     ReadGlobalSetting(Settings::values.bg_red);
     ReadGlobalSetting(Settings::values.bg_green);
     ReadGlobalSetting(Settings::values.bg_blue);
@@ -1358,6 +1360,10 @@ void Config::SaveRendererValues() {
                  Settings::values.nvdec_emulation.UsingGlobal());
     WriteGlobalSetting(Settings::values.accelerate_astc);
     WriteGlobalSetting(Settings::values.async_astc);
+    WriteSetting(QString::fromStdString(Settings::values.astc_recompression.GetLabel()),
+                 static_cast<u32>(Settings::values.astc_recompression.GetValue(global)),
+                 static_cast<u32>(Settings::values.astc_recompression.GetDefault()),
+                 Settings::values.astc_recompression.UsingGlobal());
     WriteGlobalSetting(Settings::values.use_reactive_flushing);
     WriteSetting(QString::fromStdString(Settings::values.shader_backend.GetLabel()),
                  static_cast<u32>(Settings::values.shader_backend.GetValue(global)),
@@ -1366,6 +1372,7 @@ void Config::SaveRendererValues() {
     WriteGlobalSetting(Settings::values.use_asynchronous_shaders);
     WriteGlobalSetting(Settings::values.use_fast_gpu_time);
     WriteGlobalSetting(Settings::values.use_vulkan_driver_pipeline_cache);
+    WriteGlobalSetting(Settings::values.enable_compute_pipelines);
     WriteGlobalSetting(Settings::values.bg_red);
     WriteGlobalSetting(Settings::values.bg_green);
     WriteGlobalSetting(Settings::values.bg_blue);

src/yuzu/configuration/config.h

@@ -208,3 +208,4 @@ Q_DECLARE_METATYPE(Settings::ScalingFilter);
 Q_DECLARE_METATYPE(Settings::AntiAliasing);
 Q_DECLARE_METATYPE(Settings::RendererBackend);
 Q_DECLARE_METATYPE(Settings::ShaderBackend);
+Q_DECLARE_METATYPE(Settings::AstcRecompression);

src/yuzu/configuration/configure_dialog.cpp

@@ -36,8 +36,9 @@ ConfigureDialog::ConfigureDialog(QWidget* parent, HotkeyRegistry& registry_,
       debug_tab_tab{std::make_unique<ConfigureDebugTab>(system_, this)},
       filesystem_tab{std::make_unique<ConfigureFilesystem>(this)},
       general_tab{std::make_unique<ConfigureGeneral>(system_, this)},
-      graphics_tab{std::make_unique<ConfigureGraphics>(system_, this)},
       graphics_advanced_tab{std::make_unique<ConfigureGraphicsAdvanced>(system_, this)},
+      graphics_tab{std::make_unique<ConfigureGraphics>(
+          system_, [&]() { graphics_advanced_tab->ExposeComputeOption(); }, this)},
       hotkeys_tab{std::make_unique<ConfigureHotkeys>(system_.HIDCore(), this)},
       input_tab{std::make_unique<ConfigureInput>(system_, this)},
       network_tab{std::make_unique<ConfigureNetwork>(system_, this)},

src/yuzu/configuration/configure_dialog.h

@@ -72,8 +72,8 @@ private:
     std::unique_ptr<ConfigureDebugTab> debug_tab_tab;
     std::unique_ptr<ConfigureFilesystem> filesystem_tab;
     std::unique_ptr<ConfigureGeneral> general_tab;
-    std::unique_ptr<ConfigureGraphics> graphics_tab;
     std::unique_ptr<ConfigureGraphicsAdvanced> graphics_advanced_tab;
+    std::unique_ptr<ConfigureGraphics> graphics_tab;
     std::unique_ptr<ConfigureHotkeys> hotkeys_tab;
     std::unique_ptr<ConfigureInput> input_tab;
     std::unique_ptr<ConfigureNetwork> network_tab;

src/yuzu/configuration/configure_graphics.cpp

@@ -2,9 +2,11 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 // Include this early to include Vulkan headers how we want to
+#include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
 #include <algorithm>
+#include <functional>
 #include <iosfwd>
 #include <iterator>
 #include <string>
@@ -74,8 +76,11 @@ static constexpr Settings::VSyncMode PresentModeToSetting(VkPresentModeKHR mode)
     }
 }
 
-ConfigureGraphics::ConfigureGraphics(const Core::System& system_, QWidget* parent)
-    : QWidget(parent), ui{std::make_unique<Ui::ConfigureGraphics>()}, system{system_} {
+ConfigureGraphics::ConfigureGraphics(const Core::System& system_,
+                                     const std::function<void()>& expose_compute_option_,
+                                     QWidget* parent)
+    : QWidget(parent), ui{std::make_unique<Ui::ConfigureGraphics>()},
+      expose_compute_option{expose_compute_option_}, system{system_} {
     vulkan_device = Settings::values.vulkan_device.GetValue();
     RetrieveVulkanDevices();
 
@@ -513,8 +518,7 @@ void ConfigureGraphics::RetrieveVulkanDevices() try {
     const Common::DynamicLibrary library = OpenLibrary();
     const vk::Instance instance = CreateInstance(library, dld, VK_API_VERSION_1_1, wsi.type);
     const std::vector<VkPhysicalDevice> physical_devices = instance.EnumeratePhysicalDevices();
-    vk::SurfaceKHR surface = //< needed to view present modes for a device
-        CreateSurface(instance, wsi);
+    vk::SurfaceKHR surface = CreateSurface(instance, wsi);
 
     vulkan_devices.clear();
     vulkan_devices.reserve(physical_devices.size());
@@ -527,6 +531,17 @@ void ConfigureGraphics::RetrieveVulkanDevices() try {
             physical_device.GetSurfacePresentModesKHR(*surface);
         vulkan_devices.push_back(QString::fromStdString(name));
         device_present_modes.push_back(present_modes);
+
+        VkPhysicalDeviceDriverProperties driver_properties{};
+        driver_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES;
+        driver_properties.pNext = nullptr;
+        VkPhysicalDeviceProperties2 properties{};
+        properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2_KHR;
+        properties.pNext = &driver_properties;
+        dld.vkGetPhysicalDeviceProperties2(physical_device, &properties);
+        if (driver_properties.driverID == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS) {
+            expose_compute_option();
+        }
     }
 } catch (const Vulkan::vk::Exception& exception) {
     LOG_ERROR(Frontend, "Failed to enumerate devices with error: {}", exception.what());

src/yuzu/configuration/configure_graphics.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <functional>
 #include <memory>
 #include <vector>
 #include <QColor>
@@ -37,7 +38,9 @@ class ConfigureGraphics : public QWidget {
     Q_OBJECT
 
 public:
-    explicit ConfigureGraphics(const Core::System& system_, QWidget* parent = nullptr);
+    explicit ConfigureGraphics(const Core::System& system_,
+                               const std::function<void()>& expose_compute_option_,
+                               QWidget* parent = nullptr);
     ~ConfigureGraphics() override;
 
     void ApplyConfiguration();
@@ -81,6 +84,7 @@ private:
                                       // selection in the combobox
     u32 vulkan_device{};
     Settings::ShaderBackend shader_backend{};
+    const std::function<void()>& expose_compute_option;
 
     const Core::System& system;
 };

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -15,6 +15,8 @@ ConfigureGraphicsAdvanced::ConfigureGraphicsAdvanced(const Core::System& system_
     SetupPerGameUI();
 
     SetConfiguration();
+
+    ui->enable_compute_pipelines_checkbox->setVisible(false);
 }
 
 ConfigureGraphicsAdvanced::~ConfigureGraphicsAdvanced() = default;
@@ -25,8 +27,10 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
     ui->async_present->setEnabled(runtime_lock);
     ui->renderer_force_max_clock->setEnabled(runtime_lock);
     ui->async_astc->setEnabled(runtime_lock);
+    ui->astc_recompression_combobox->setEnabled(runtime_lock);
     ui->use_asynchronous_shaders->setEnabled(runtime_lock);
     ui->anisotropic_filtering_combobox->setEnabled(runtime_lock);
+    ui->enable_compute_pipelines_checkbox->setEnabled(runtime_lock);
 
     ui->async_present->setChecked(Settings::values.async_presentation.GetValue());
     ui->renderer_force_max_clock->setChecked(Settings::values.renderer_force_max_clock.GetValue());
@@ -36,20 +40,28 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time.GetValue());
     ui->use_vulkan_driver_pipeline_cache->setChecked(
         Settings::values.use_vulkan_driver_pipeline_cache.GetValue());
+    ui->enable_compute_pipelines_checkbox->setChecked(
+        Settings::values.enable_compute_pipelines.GetValue());
 
     if (Settings::IsConfiguringGlobal()) {
         ui->gpu_accuracy->setCurrentIndex(
             static_cast<int>(Settings::values.gpu_accuracy.GetValue()));
         ui->anisotropic_filtering_combobox->setCurrentIndex(
             Settings::values.max_anisotropy.GetValue());
+        ui->astc_recompression_combobox->setCurrentIndex(
+            static_cast<int>(Settings::values.astc_recompression.GetValue()));
     } else {
         ConfigurationShared::SetPerGameSetting(ui->gpu_accuracy, &Settings::values.gpu_accuracy);
         ConfigurationShared::SetPerGameSetting(ui->anisotropic_filtering_combobox,
                                                &Settings::values.max_anisotropy);
+        ConfigurationShared::SetPerGameSetting(ui->astc_recompression_combobox,
+                                               &Settings::values.astc_recompression);
         ConfigurationShared::SetHighlight(ui->label_gpu_accuracy,
                                           !Settings::values.gpu_accuracy.UsingGlobal());
         ConfigurationShared::SetHighlight(ui->af_label,
                                           !Settings::values.max_anisotropy.UsingGlobal());
+        ConfigurationShared::SetHighlight(ui->label_astc_recompression,
+                                          !Settings::values.astc_recompression.UsingGlobal());
     }
 }
 
@@ -66,6 +78,8 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
                                              ui->use_reactive_flushing, use_reactive_flushing);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.async_astc, ui->async_astc,
                                              async_astc);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.astc_recompression,
+                                             ui->astc_recompression_combobox);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_asynchronous_shaders,
                                              ui->use_asynchronous_shaders,
                                              use_asynchronous_shaders);
@@ -74,6 +88,9 @@ void ConfigureGraphicsAdvanced::ApplyConfiguration() {
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_vulkan_driver_pipeline_cache,
                                              ui->use_vulkan_driver_pipeline_cache,
                                              use_vulkan_driver_pipeline_cache);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.enable_compute_pipelines,
+                                             ui->enable_compute_pipelines_checkbox,
+                                             enable_compute_pipelines);
 }
 
 void ConfigureGraphicsAdvanced::changeEvent(QEvent* event) {
@@ -97,6 +114,8 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
             Settings::values.renderer_force_max_clock.UsingGlobal());
         ui->use_reactive_flushing->setEnabled(Settings::values.use_reactive_flushing.UsingGlobal());
         ui->async_astc->setEnabled(Settings::values.async_astc.UsingGlobal());
+        ui->astc_recompression_combobox->setEnabled(
+            Settings::values.astc_recompression.UsingGlobal());
         ui->use_asynchronous_shaders->setEnabled(
             Settings::values.use_asynchronous_shaders.UsingGlobal());
         ui->use_fast_gpu_time->setEnabled(Settings::values.use_fast_gpu_time.UsingGlobal());
@@ -104,6 +123,8 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
             Settings::values.use_vulkan_driver_pipeline_cache.UsingGlobal());
         ui->anisotropic_filtering_combobox->setEnabled(
             Settings::values.max_anisotropy.UsingGlobal());
+        ui->enable_compute_pipelines_checkbox->setEnabled(
+            Settings::values.enable_compute_pipelines.UsingGlobal());
 
         return;
     }
@@ -125,10 +146,20 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
     ConfigurationShared::SetColoredTristate(ui->use_vulkan_driver_pipeline_cache,
                                             Settings::values.use_vulkan_driver_pipeline_cache,
                                             use_vulkan_driver_pipeline_cache);
+    ConfigurationShared::SetColoredTristate(ui->enable_compute_pipelines_checkbox,
+                                            Settings::values.enable_compute_pipelines,
+                                            enable_compute_pipelines);
     ConfigurationShared::SetColoredComboBox(
         ui->gpu_accuracy, ui->label_gpu_accuracy,
         static_cast<int>(Settings::values.gpu_accuracy.GetValue(true)));
     ConfigurationShared::SetColoredComboBox(
         ui->anisotropic_filtering_combobox, ui->af_label,
         static_cast<int>(Settings::values.max_anisotropy.GetValue(true)));
+    ConfigurationShared::SetColoredComboBox(
+        ui->astc_recompression_combobox, ui->label_astc_recompression,
+        static_cast<int>(Settings::values.astc_recompression.GetValue(true)));
+}
+
+void ConfigureGraphicsAdvanced::ExposeComputeOption() {
+    ui->enable_compute_pipelines_checkbox->setVisible(true);
 }

src/yuzu/configuration/configure_graphics_advanced.h

@@ -28,6 +28,8 @@ public:
     void ApplyConfiguration();
     void SetConfiguration();
 
+    void ExposeComputeOption();
+
 private:
     void changeEvent(QEvent* event) override;
     void RetranslateUI();
@@ -44,6 +46,7 @@ private:
     ConfigurationShared::CheckState use_asynchronous_shaders;
     ConfigurationShared::CheckState use_fast_gpu_time;
     ConfigurationShared::CheckState use_vulkan_driver_pipeline_cache;
+    ConfigurationShared::CheckState enable_compute_pipelines;
 
     const Core::System& system;
 };

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -69,6 +69,50 @@
           </layout>
          </widget>
         </item>
+        <item>
+         <widget class="QWidget" name="astc_recompression_layout" native="true">
+          <layout class="QHBoxLayout" name="horizontalLayout_3">
+            <property name="leftMargin">
+              <number>0</number>
+            </property>
+            <property name="topMargin">
+              <number>0</number>
+            </property>
+            <property name="rightMargin">
+              <number>0</number>
+            </property>
+            <property name="bottomMargin">
+              <number>0</number>
+            </property>
+            <item>
+              <widget class="QLabel" name="label_astc_recompression">
+              <property name="text">
+                <string>ASTC recompression:</string>
+              </property>
+              </widget>
+            </item>
+            <item>
+              <widget class="QComboBox" name="astc_recompression_combobox">
+                <item>
+                <property name="text">
+                  <string>Uncompressed (Best quality)</string>
+                </property>
+                </item>
+                <item>
+                <property name="text">
+                  <string>BC1 (Low quality)</string>
+                </property>
+                </item>
+                <item>
+                <property name="text">
+                  <string>BC3 (Medium quality)</string>
+                </property>
+                </item>
+              </widget>
+            </item>
+          </layout>
+         </widget>
+        </item>
         <item>
          <widget class="QCheckBox" name="async_present">
           <property name="text">
@@ -136,6 +180,17 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="enable_compute_pipelines_checkbox">
+          <property name="toolTip">
+           <string>Enable compute pipelines, required by some games. This setting only exists for Intel proprietary drivers, and may crash if enabled.
+Compute pipelines are always enabled on all other drivers.</string>
+          </property>
+          <property name="text">
+           <string>Enable Compute Pipelines (Intel Vulkan only)</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <widget class="QWidget" name="af_layout" native="true">
           <layout class="QHBoxLayout" name="horizontalLayout_1">

src/yuzu/configuration/configure_hotkeys.cpp

@@ -48,7 +48,9 @@ ConfigureHotkeys::ConfigureHotkeys(Core::HID::HIDCore& hid_core, QWidget* parent
 
     connect(poll_timer.get(), &QTimer::timeout, [this] {
         const auto buttons = controller->GetNpadButtons();
-        if (buttons.raw != Core::HID::NpadButton::None) {
+        const auto home_pressed = controller->GetHomeButtons().home != 0;
+        const auto capture_pressed = controller->GetCaptureButtons().capture != 0;
+        if (home_pressed || capture_pressed) {
             SetPollingResult(buttons.raw, false);
             return;
         }
@@ -154,8 +156,10 @@ void ConfigureHotkeys::ConfigureController(QModelIndex index) {
             model->setData(index, previous_key);
             return;
         }
-
-        const QString button_string = tr("Home+%1").arg(GetButtonName(button));
+        const auto home_pressed = this->controller->GetHomeButtons().home != 0;
+        const auto capture_pressed = this->controller->GetCaptureButtons().capture != 0;
+        const QString button_string =
+            GetButtonCombinationName(button, home_pressed, capture_pressed);
 
         const auto [key_sequence_used, used_action] = IsUsedControllerKey(button_string);
 
@@ -174,72 +178,83 @@ void ConfigureHotkeys::ConfigureController(QModelIndex index) {
     poll_timer->start(200);     // Check for new inputs every 200ms
     // We need to disable configuration to be able to read npad buttons
     controller->DisableConfiguration();
-    controller->DisableSystemButtons();
 }
 
 void ConfigureHotkeys::SetPollingResult(Core::HID::NpadButton button, const bool cancel) {
     timeout_timer->stop();
     poll_timer->stop();
+    (*input_setter)(button, cancel);
     // Re-Enable configuration
     controller->EnableConfiguration();
-    controller->EnableSystemButtons();
-
-    (*input_setter)(button, cancel);
 
     input_setter = std::nullopt;
 }
 
-QString ConfigureHotkeys::GetButtonName(Core::HID::NpadButton button) const {
+QString ConfigureHotkeys::GetButtonCombinationName(Core::HID::NpadButton button,
+                                                   const bool home = false,
+                                                   const bool capture = false) const {
     Core::HID::NpadButtonState state{button};
+    QString button_combination;
+    if (home) {
+        button_combination.append(QStringLiteral("Home+"));
+    }
+    if (capture) {
+        button_combination.append(QStringLiteral("Screenshot+"));
+    }
     if (state.a) {
-        return QStringLiteral("A");
+        button_combination.append(QStringLiteral("A+"));
     }
     if (state.b) {
-        return QStringLiteral("B");
+        button_combination.append(QStringLiteral("B+"));
     }
     if (state.x) {
-        return QStringLiteral("X");
+        button_combination.append(QStringLiteral("X+"));
     }
     if (state.y) {
-        return QStringLiteral("Y");
+        button_combination.append(QStringLiteral("Y+"));
     }
     if (state.l || state.right_sl || state.left_sl) {
-        return QStringLiteral("L");
+        button_combination.append(QStringLiteral("L+"));
     }
     if (state.r || state.right_sr || state.left_sr) {
-        return QStringLiteral("R");
+        button_combination.append(QStringLiteral("R+"));
     }
     if (state.zl) {
-        return QStringLiteral("ZL");
+        button_combination.append(QStringLiteral("ZL+"));
     }
     if (state.zr) {
-        return QStringLiteral("ZR");
+        button_combination.append(QStringLiteral("ZR+"));
     }
     if (state.left) {
-        return QStringLiteral("Dpad_Left");
+        button_combination.append(QStringLiteral("Dpad_Left+"));
     }
     if (state.right) {
-        return QStringLiteral("Dpad_Right");
+        button_combination.append(QStringLiteral("Dpad_Right+"));
     }
     if (state.up) {
-        return QStringLiteral("Dpad_Up");
+        button_combination.append(QStringLiteral("Dpad_Up+"));
     }
     if (state.down) {
-        return QStringLiteral("Dpad_Down");
+        button_combination.append(QStringLiteral("Dpad_Down+"));
     }
     if (state.stick_l) {
-        return QStringLiteral("Left_Stick");
+        button_combination.append(QStringLiteral("Left_Stick+"));
     }
     if (state.stick_r) {
-        return QStringLiteral("Right_Stick");
+        button_combination.append(QStringLiteral("Right_Stick+"));
     }
     if (state.minus) {
-        return QStringLiteral("Minus");
+        button_combination.append(QStringLiteral("Minus+"));
     }
     if (state.plus) {
-        return QStringLiteral("Plus");
+        button_combination.append(QStringLiteral("Plus+"));
+    }
+    if (button_combination.isEmpty()) {
+        return tr("Invalid");
+    } else {
+        button_combination.chop(1);
+        return button_combination;
     }
-    return tr("Invalid");
 }
 
 std::pair<bool, QString> ConfigureHotkeys::IsUsedKey(QKeySequence key_sequence) const {

src/yuzu/configuration/configure_hotkeys.h

@@ -59,7 +59,7 @@ private:
     QStandardItemModel* model;
 
     void SetPollingResult(Core::HID::NpadButton button, bool cancel);
-    QString GetButtonName(Core::HID::NpadButton button) const;
+    QString GetButtonCombinationName(Core::HID::NpadButton button, bool home, bool capture) const;
     Core::HID::EmulatedController* controller;
     std::unique_ptr<QTimer> timeout_timer;
     std::unique_ptr<QTimer> poll_timer;

src/yuzu/configuration/configure_per_game.cpp

@@ -48,8 +48,9 @@ ConfigurePerGame::ConfigurePerGame(QWidget* parent, u64 title_id_, const std::st
     audio_tab = std::make_unique<ConfigureAudio>(system_, this);
     cpu_tab = std::make_unique<ConfigureCpu>(system_, this);
     general_tab = std::make_unique<ConfigureGeneral>(system_, this);
-    graphics_tab = std::make_unique<ConfigureGraphics>(system_, this);
     graphics_advanced_tab = std::make_unique<ConfigureGraphicsAdvanced>(system_, this);
+    graphics_tab = std::make_unique<ConfigureGraphics>(
+        system_, [&]() { graphics_advanced_tab->ExposeComputeOption(); }, this);
     input_tab = std::make_unique<ConfigureInputPerGame>(system_, game_config.get(), this);
     system_tab = std::make_unique<ConfigureSystem>(system_, this);
 

src/yuzu/configuration/configure_per_game.h

@@ -75,8 +75,8 @@ private:
     std::unique_ptr<ConfigureAudio> audio_tab;
     std::unique_ptr<ConfigureCpu> cpu_tab;
     std::unique_ptr<ConfigureGeneral> general_tab;
-    std::unique_ptr<ConfigureGraphics> graphics_tab;
     std::unique_ptr<ConfigureGraphicsAdvanced> graphics_advanced_tab;
+    std::unique_ptr<ConfigureGraphics> graphics_tab;
     std::unique_ptr<ConfigureInputPerGame> input_tab;
     std::unique_ptr<ConfigureSystem> system_tab;
 };

src/yuzu/main.cpp

@@ -1164,7 +1164,8 @@ void GMainWindow::InitializeRecentFileMenuActions() {
     UpdateRecentFiles();
 }
 
-void GMainWindow::LinkActionShortcut(QAction* action, const QString& action_name) {
+void GMainWindow::LinkActionShortcut(QAction* action, const QString& action_name,
+                                     const bool tas_allowed) {
     static const QString main_window = QStringLiteral("Main Window");
     action->setShortcut(hotkey_registry.GetKeySequence(main_window, action_name));
     action->setShortcutContext(hotkey_registry.GetShortcutContext(main_window, action_name));
@@ -1176,7 +1177,14 @@ void GMainWindow::LinkActionShortcut(QAction* action, const QString& action_name
     const auto* controller_hotkey =
         hotkey_registry.GetControllerHotkey(main_window, action_name, controller);
     connect(
-        controller_hotkey, &ControllerShortcut::Activated, this, [action] { action->trigger(); },
+        controller_hotkey, &ControllerShortcut::Activated, this,
+        [action, tas_allowed, this] {
+            auto [tas_status, current_tas_frame, total_tas_frames] =
+                input_subsystem->GetTas()->GetStatus();
+            if (tas_allowed || tas_status == InputCommon::TasInput::TasState::Stopped) {
+                action->trigger();
+            }
+        },
         Qt::QueuedConnection);
 }
 
@@ -1193,9 +1201,9 @@ void GMainWindow::InitializeHotkeys() {
     LinkActionShortcut(ui->action_Show_Status_Bar, QStringLiteral("Toggle Status Bar"));
     LinkActionShortcut(ui->action_Fullscreen, QStringLiteral("Fullscreen"));
     LinkActionShortcut(ui->action_Capture_Screenshot, QStringLiteral("Capture Screenshot"));
-    LinkActionShortcut(ui->action_TAS_Start, QStringLiteral("TAS Start/Stop"));
-    LinkActionShortcut(ui->action_TAS_Record, QStringLiteral("TAS Record"));
-    LinkActionShortcut(ui->action_TAS_Reset, QStringLiteral("TAS Reset"));
+    LinkActionShortcut(ui->action_TAS_Start, QStringLiteral("TAS Start/Stop"), true);
+    LinkActionShortcut(ui->action_TAS_Record, QStringLiteral("TAS Record"), true);
+    LinkActionShortcut(ui->action_TAS_Reset, QStringLiteral("TAS Reset"), true);
 
     static const QString main_window = QStringLiteral("Main Window");
     const auto connect_shortcut = [&]<typename Fn>(const QString& action_name, const Fn& function) {

src/yuzu/main.h

@@ -214,7 +214,8 @@ public slots:
 
 private:
     /// Updates an action's shortcut and text to reflect an updated hotkey from the hotkey registry.
-    void LinkActionShortcut(QAction* action, const QString& action_name);
+    void LinkActionShortcut(QAction* action, const QString& action_name,
+                            const bool tas_allowed = false);
 
     void RegisterMetaTypes();
 

src/yuzu_cmd/config.cpp

@@ -318,6 +318,7 @@ void Config::ReadValues() {
     ReadSetting("Renderer", Settings::values.nvdec_emulation);
     ReadSetting("Renderer", Settings::values.accelerate_astc);
     ReadSetting("Renderer", Settings::values.async_astc);
+    ReadSetting("Renderer", Settings::values.astc_recompression);
     ReadSetting("Renderer", Settings::values.use_fast_gpu_time);
     ReadSetting("Renderer", Settings::values.use_vulkan_driver_pipeline_cache);
 

src/yuzu_cmd/default_ini.h

@@ -360,6 +360,10 @@ accelerate_astc =
 # 0 (default): Off, 1: On
 async_astc =
 
+# Recompress ASTC textures to a different format.
+# 0 (default): Uncompressed, 1: BC1 (Low quality), 2: BC3: (Medium quality)
+async_astc =
+
 # Turns on the speed limiter, which will limit the emulation speed to the desired speed limit value
 # 0: Off, 1: On (default)
 use_speed_limit =