Merge branch 'main' into qt_save

.github/workflows/build.yml

@@ -76,18 +76,13 @@ jobs:
             ${{ env.cache-name }}-
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-sdl-cache-cmake-build
       with:
         append-timestamp: false
         key: ${{ env.cache-name }}-${{ hashFiles('**/CMakeLists.txt', 'cmake/**') }}
 
-    - name: Setup VS Environment
-      uses: ilammy/msvc-dev-cmd@v1.13.0
-      with:
-        arch: amd64
-
     - name: Configure CMake
       run: cmake --fresh -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_INTERPROCEDURAL_OPTIMIZATION_RELEASE=ON -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache
 
@@ -111,7 +106,7 @@ jobs:
     - name: Setup Qt
       uses: jurplel/install-qt-action@v4
       with:
-        version: 6.9.0
+        version: 6.9.1
         host: windows
         target: desktop
         arch: win64_msvc2022_64
@@ -130,18 +125,13 @@ jobs:
             ${{ env.cache-name }}-
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-qt-cache-cmake-build
       with:
         append-timestamp: false
         key: ${{ env.cache-name }}-${{ hashFiles('**/CMakeLists.txt', 'cmake/**') }}
 
-    - name: Setup VS Environment
-      uses: ilammy/msvc-dev-cmd@v1.13.0
-      with:
-        arch: amd64
-
     - name: Configure CMake
       run: cmake --fresh -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DENABLE_QT_GUI=ON -DENABLE_UPDATER=ON -DCMAKE_INTERPROCEDURAL_OPTIMIZATION_RELEASE=ON -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache
 
@@ -186,7 +176,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{runner.os}}-sdl-cache-cmake-build
       with:
@@ -228,7 +218,7 @@ jobs:
     - name: Setup Qt
       uses: jurplel/install-qt-action@v4
       with:
-        version: 6.9.0
+        version: 6.9.1
         host: mac
         target: desktop
         arch: clang_64
@@ -247,7 +237,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{runner.os}}-qt-cache-cmake-build
       with:
@@ -301,7 +291,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-sdl-cache-cmake-build
       with:
@@ -362,7 +352,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-qt-cache-cmake-build
       with:
@@ -409,7 +399,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-sdl-gcc-cache-cmake-build
       with:
@@ -445,7 +435,7 @@ jobs:
             ${{ env.cache-name }}- 
 
     - name: Cache CMake Build
-      uses: hendrikmuhs/ccache-action@v1.2.17
+      uses: hendrikmuhs/ccache-action@v1.2.18
       env:
           cache-name: ${{ runner.os }}-qt-gcc-cache-cmake-build
       with:

src/core/libraries/videoout/video_out.cpp

@@ -282,7 +282,12 @@ s32 PS4_SYSV_ABI sceVideoOutGetVblankStatus(int handle, SceVideoOutVblankStatus*
 
 s32 PS4_SYSV_ABI sceVideoOutGetResolutionStatus(s32 handle, SceVideoOutResolutionStatus* status) {
     LOG_INFO(Lib_VideoOut, "called");
-    *status = driver->GetPort(handle)->resolution;
+    auto* port = driver->GetPort(handle);
+    if (!port || !port->is_open) {
+        return ORBIS_VIDEO_OUT_ERROR_INVALID_HANDLE;
+    }
+
+    *status = port->resolution;
     return ORBIS_OK;
 }
 

src/core/libraries/zlib/zlib.cpp

@@ -51,7 +51,7 @@ void ZlibTaskThread(const std::stop_token& stop) {
             if (!task_queue_cv.wait(lock, stop, [&] { return !task_queue.empty(); })) {
                 break;
             }
-            task = task_queue.back();
+            task = task_queue.front();
             task_queue.pop();
         }
 
@@ -136,7 +136,7 @@ s32 PS4_SYSV_ABI sceZlibWaitForDone(u64* request_id, const u32* timeout) {
         } else {
             done_queue_cv.wait(lock, pred);
         }
-        *request_id = done_queue.back();
+        *request_id = done_queue.front();
         done_queue.pop();
     }
     return ORBIS_OK;

src/shader_recompiler/backend/spirv/emit_spirv_instructions.h

@@ -372,6 +372,7 @@ Id EmitBitCount64(EmitContext& ctx, Id value);
 Id EmitBitwiseNot32(EmitContext& ctx, Id value);
 Id EmitFindSMsb32(EmitContext& ctx, Id value);
 Id EmitFindUMsb32(EmitContext& ctx, Id value);
+Id EmitFindUMsb64(EmitContext& ctx, Id value);
 Id EmitFindILsb32(EmitContext& ctx, Id value);
 Id EmitFindILsb64(EmitContext& ctx, Id value);
 Id EmitSMin32(EmitContext& ctx, Id a, Id b);

src/shader_recompiler/backend/spirv/emit_spirv_integer.cpp

@@ -229,6 +229,20 @@ Id EmitFindUMsb32(EmitContext& ctx, Id value) {
     return ctx.OpFindUMsb(ctx.U32[1], value);
 }
 
+Id EmitFindUMsb64(EmitContext& ctx, Id value) {
+    // Vulkan restricts some bitwise operations to 32-bit only, so decompose into
+    // two 32-bit values and select the correct result.
+    const Id unpacked{ctx.OpBitcast(ctx.U32[2], value)};
+    const Id hi{ctx.OpCompositeExtract(ctx.U32[1], unpacked, 1U)};
+    const Id lo{ctx.OpCompositeExtract(ctx.U32[1], unpacked, 0U)};
+    const Id hi_msb{ctx.OpFindUMsb(ctx.U32[1], hi)};
+    const Id lo_msb{ctx.OpFindUMsb(ctx.U32[1], lo)};
+    const Id found_hi{ctx.OpINotEqual(ctx.U1[1], hi_msb, ctx.ConstU32(u32(-1)))};
+    const Id shifted_hi{ctx.OpIAdd(ctx.U32[1], hi_msb, ctx.ConstU32(32u))};
+    // value == 0 case is checked in IREmitter
+    return ctx.OpSelect(ctx.U32[1], found_hi, shifted_hi, lo_msb);
+}
+
 Id EmitFindILsb32(EmitContext& ctx, Id value) {
     return ctx.OpFindILsb(ctx.U32[1], value);
 }

src/shader_recompiler/frontend/translate/scalar_alu.cpp

@@ -114,6 +114,8 @@ void Translator::EmitScalarAlu(const GcnInst& inst) {
             return S_FF1_I32_B64(inst);
         case Opcode::S_FLBIT_I32_B32:
             return S_FLBIT_I32_B32(inst);
+        case Opcode::S_FLBIT_I32_B64:
+            return S_FLBIT_I32_B64(inst);
         case Opcode::S_BITSET0_B32:
             return S_BITSET_B32(inst, 0);
         case Opcode::S_BITSET1_B32:
@@ -686,6 +688,17 @@ void Translator::S_FLBIT_I32_B32(const GcnInst& inst) {
     SetDst(inst.dst[0], IR::U32{ir.Select(cond, pos_from_left, ir.Imm32(~0U))});
 }
 
+void Translator::S_FLBIT_I32_B64(const GcnInst& inst) {
+    const IR::U64 src0{GetSrc64(inst.src[0])};
+    // Gcn wants the MSB position counting from the left, but SPIR-V counts from the rightmost (LSB)
+    // position
+    const IR::U32 msb_pos = ir.FindUMsb(src0);
+    const IR::U32 pos_from_left = ir.ISub(ir.Imm32(63), msb_pos);
+    // Select 0xFFFFFFFF if src0 was 0
+    const IR::U1 cond = ir.INotEqual(src0, ir.Imm64(u64(0u)));
+    SetDst(inst.dst[0], IR::U32{ir.Select(cond, pos_from_left, ir.Imm32(~0U))});
+}
+
 void Translator::S_BITSET_B32(const GcnInst& inst, u32 bit_value) {
     const IR::U32 old_value{GetSrc(inst.dst[0])};
     const IR::U32 offset{ir.BitFieldExtract(GetSrc(inst.src[0]), ir.Imm32(0U), ir.Imm32(5U))};

src/shader_recompiler/frontend/translate/translate.h

@@ -121,6 +121,7 @@ public:
     void S_FF1_I32_B32(const GcnInst& inst);
     void S_FF1_I32_B64(const GcnInst& inst);
     void S_FLBIT_I32_B32(const GcnInst& inst);
+    void S_FLBIT_I32_B64(const GcnInst& inst);
     void S_BITSET_B32(const GcnInst& inst, u32 bit_value);
     void S_GETPC_B64(u32 pc, const GcnInst& inst);
     void S_SAVEEXEC_B64(NegateMode negate, bool is_or, const GcnInst& inst);

src/shader_recompiler/ir/ir_emitter.cpp

@@ -1546,8 +1546,15 @@ U32 IREmitter::FindSMsb(const U32& value) {
     return Inst<U32>(Opcode::FindSMsb32, value);
 }
 
-U32 IREmitter::FindUMsb(const U32& value) {
-    return Inst<U32>(Opcode::FindUMsb32, value);
+U32 IREmitter::FindUMsb(const U32U64& value) {
+    switch (value.Type()) {
+    case Type::U32:
+        return Inst<U32>(Opcode::FindUMsb32, value);
+    case Type::U64:
+        return Inst<U32>(Opcode::FindUMsb64, value);
+    default:
+        ThrowInvalidType(value.Type());
+    }
 }
 
 U32 IREmitter::FindILsb(const U32U64& value) {

src/shader_recompiler/ir/ir_emitter.h

@@ -266,7 +266,7 @@ public:
     [[nodiscard]] U32 BitwiseNot(const U32& value);
 
     [[nodiscard]] U32 FindSMsb(const U32& value);
-    [[nodiscard]] U32 FindUMsb(const U32& value);
+    [[nodiscard]] U32 FindUMsb(const U32U64& value);
     [[nodiscard]] U32 FindILsb(const U32U64& value);
     [[nodiscard]] U32 SMin(const U32& a, const U32& b);
     [[nodiscard]] U32 UMin(const U32& a, const U32& b);

src/shader_recompiler/ir/opcodes.inc

@@ -349,6 +349,7 @@ OPCODE(BitwiseNot32,                                        U32,            U32,
 
 OPCODE(FindSMsb32,                                          U32,            U32,                                                                            )
 OPCODE(FindUMsb32,                                          U32,            U32,                                                                            )
+OPCODE(FindUMsb64,                                          U32,            U64,                                                                            )
 OPCODE(FindILsb32,                                          U32,            U32,                                                                            )
 OPCODE(FindILsb64,                                          U32,            U64,                                                                            )
 OPCODE(SMin32,                                              U32,            U32,            U32,                                                            )

src/video_core/amdgpu/liverpool.cpp

@@ -765,6 +765,19 @@ Liverpool::Task Liverpool::ProcessGraphics(std::span<const u32> dcb, std::span<c
                 LOG_WARNING(Render_Vulkan, "Unimplemented IT_GET_LOD_STATS");
                 break;
             }
+            case PM4ItOpcode::CondExec: {
+                const auto* cond_exec = reinterpret_cast<const PM4CmdCondExec*>(header);
+                if (cond_exec->command.Value() != 0) {
+                    LOG_WARNING(Render, "IT_COND_EXEC used a reserved command");
+                }
+                const auto skip = *cond_exec->Address() == false;
+                if (skip) {
+                    dcb = NextPacket(dcb,
+                                     header->type3.NumWords() + 1 + cond_exec->exec_count.Value());
+                    continue;
+                }
+                break;
+            }
             default:
                 UNREACHABLE_MSG("Unknown PM4 type 3 opcode {:#x} with count {}",
                                 static_cast<u32>(opcode), count);

src/video_core/amdgpu/pm4_cmds.h

@@ -1159,4 +1159,25 @@ struct PM4CmdMemSemaphore {
     }
 };
 
+struct PM4CmdCondExec {
+    PM4Type3Header header;
+    union {
+        BitField<2, 30, u32> bool_addr_lo; ///< low 32 address bits for the block in memory from
+                                           ///< where the CP will fetch the condition
+    };
+    union {
+        BitField<0, 16, u32> bool_addr_hi; ///< high address bits for the condition
+        BitField<28, 4, u32> command;
+    };
+    union {
+        BitField<0, 14, u32> exec_count; ///< Number of DWords that the CP will skip
+                                         ///< if bool pointed to is zero
+    };
+
+    bool* Address() const {
+        return std::bit_cast<bool*>(u64(bool_addr_hi.Value()) << 32 | u64(bool_addr_lo.Value())
+                                                                          << 2);
+    }
+};
+
 } // namespace AmdGpu