early-access version 2637

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 2636.
+This is the source code for early-access 2637.
 
 ## Legal Notice
 

externals/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.cpp.inc

@@ -75,6 +75,14 @@ void AxxEmitX64::EmitMemoryRead(AxxEmitContext& ctx, IR::Inst* inst) {
         return;
     }
 
+    if (ordered && bitsize == 128) {
+        // Required for atomic 128-bit loads/stores
+        ctx.reg_alloc.ScratchGpr(HostLoc::RAX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RBX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RCX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RDX);
+    }
+
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(args[0]);
     const int value_idx = bitsize == 128 ? ctx.reg_alloc.ScratchXmm().getIdx() : ctx.reg_alloc.ScratchGpr().getIdx();
 
@@ -145,6 +153,14 @@ void AxxEmitX64::EmitMemoryWrite(AxxEmitContext& ctx, IR::Inst* inst) {
         return;
     }
 
+    if (ordered && bitsize == 128) {
+        // Required for atomic 128-bit loads/stores
+        ctx.reg_alloc.ScratchGpr(HostLoc::RAX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RBX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RCX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RDX);
+    }
+
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(args[0]);
     const int value_idx = bitsize == 128
                             ? ctx.reg_alloc.UseXmm(args[1]).getIdx()
@@ -303,7 +319,15 @@ void AxxEmitX64::EmitExclusiveReadMemoryInline(AxxEmitContext& ctx, IR::Inst* in
     }
 
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    const bool ordered = IsOrdered(args[1].GetImmediateAccType());
+    const bool ordered = true;
+
+    if (ordered && bitsize == 128) {
+        // Required for atomic 128-bit loads/stores
+        ctx.reg_alloc.ScratchGpr(HostLoc::RAX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RBX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RCX);
+        ctx.reg_alloc.ScratchGpr(HostLoc::RDX);
+    }
 
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(args[0]);
     const int value_idx = bitsize == 128 ? ctx.reg_alloc.ScratchXmm().getIdx() : ctx.reg_alloc.ScratchGpr().getIdx();
@@ -370,7 +394,7 @@ void AxxEmitX64::EmitExclusiveWriteMemoryInline(AxxEmitContext& ctx, IR::Inst* i
     }
 
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-    const bool ordered = IsOrdered(args[2].GetImmediateAccType());
+    const bool ordered = true;
 
     const auto value = [&] {
         if constexpr (bitsize == 128) {

externals/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.h

@@ -201,10 +201,13 @@ template<>
 template<std::size_t bitsize>
 const void* EmitReadMemoryMov(BlockOfCode& code, int value_idx, const Xbyak::RegExp& addr, bool ordered) {
     if (ordered) {
-        if constexpr (bitsize == 128) {
+        if constexpr (bitsize != 128) {
-            code.mfence();
-        } else {
             code.xor_(Xbyak::Reg32{value_idx}, Xbyak::Reg32{value_idx});
+        } else {
+            code.xor_(eax, eax);
+            code.xor_(ebx, ebx);
+            code.xor_(ecx, ecx);
+            code.xor_(edx, edx);
         }
 
         const void* fastmem_location = code.getCurr();
@@ -215,7 +218,7 @@ const void* EmitReadMemoryMov(BlockOfCode& code, int value_idx, const Xbyak::Reg
             break;
         case 16:
             code.lock();
-            code.xadd(word[addr], Xbyak::Reg32{value_idx});
+            code.xadd(word[addr], Xbyak::Reg16{value_idx});
             break;
         case 32:
             code.lock();
@@ -226,8 +229,16 @@ const void* EmitReadMemoryMov(BlockOfCode& code, int value_idx, const Xbyak::Reg
             code.xadd(qword[addr], Xbyak::Reg64{value_idx});
             break;
         case 128:
-            // TODO (HACK): Detect CPUs where this load is not atomic
+            code.lock();
-            code.movaps(Xbyak::Xmm{value_idx}, xword[addr]);
+            code.cmpxchg16b(xword[addr]);
+            if (code.HasHostFeature(HostFeature::SSE41)) {
+                code.movq(Xbyak::Xmm{value_idx}, rax);
+                code.pinsrq(Xbyak::Xmm{value_idx}, rdx, 1);
+            } else {
+                code.movq(Xbyak::Xmm{value_idx}, rax);
+                code.movq(xmm0, rdx);
+                code.punpcklqdq(Xbyak::Xmm{value_idx}, xmm0);
+            }
             break;
         default:
             ASSERT_FALSE("Invalid bitsize");
@@ -261,6 +272,20 @@ const void* EmitReadMemoryMov(BlockOfCode& code, int value_idx, const Xbyak::Reg
 template<std::size_t bitsize>
 const void* EmitWriteMemoryMov(BlockOfCode& code, const Xbyak::RegExp& addr, int value_idx, bool ordered) {
     if (ordered) {
+        if constexpr (bitsize == 128) {
+            code.xor_(eax, eax);
+            code.xor_(edx, edx);
+            if (code.HasHostFeature(HostFeature::SSE41)) {
+                code.movq(rbx, Xbyak::Xmm{value_idx});
+                code.pextrq(rcx, Xbyak::Xmm{value_idx}, 1);
+            } else {
+                code.movaps(xmm0, Xbyak::Xmm{value_idx});
+                code.movq(rbx, xmm0);
+                code.punpckhqdq(xmm0, xmm0);
+                code.movq(rcx, xmm0);
+            }
+        }
+
         const void* fastmem_location = code.getCurr();
         switch (bitsize) {
         case 8:
@@ -275,10 +300,14 @@ const void* EmitWriteMemoryMov(BlockOfCode& code, const Xbyak::RegExp& addr, int
         case 64:
             code.xchg(qword[addr], Xbyak::Reg64{value_idx});
             break;
-        case 128:
+        case 128: {
-            code.movaps(xword[addr], Xbyak::Xmm{value_idx});
+            Xbyak::Label loop;
-            code.mfence();
+            code.L(loop);
+            code.lock();
+            code.cmpxchg16b(xword[addr]);
+            code.jnz(loop);
             break;
+        }
         default:
             ASSERT_FALSE("Invalid bitsize");
         }

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

@@ -174,7 +174,7 @@ ResultCode VfsDirectoryServiceWrapper::RenameFile(const std::string& src_path_,
     ASSERT_MSG(dest != nullptr, "Newly created file with success cannot be found.");
 
     ASSERT_MSG(dest->WriteBytes(src->ReadAllBytes()) == src->GetSize(),
-               "Could not write all of the bytes but everything else has succeded.");
+               "Could not write all of the bytes but everything else has succeeded.");
 
     if (!src->GetContainingDirectory()->DeleteFile(Common::FS::GetFilename(src_path))) {
         // TODO(DarkLordZach): Find a better error code for this

src/core/hle/service/nvdrv/nvdrv_interface.cpp

@@ -26,7 +26,7 @@ void NVDRV::Open(Kernel::HLERequestContext& ctx) {
         rb.Push<DeviceFD>(0);
         rb.PushEnum(NvResult::NotInitialized);
 
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -61,7 +61,7 @@ void NVDRV::Ioctl1(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -87,7 +87,7 @@ void NVDRV::Ioctl2(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -114,7 +114,7 @@ void NVDRV::Ioctl3(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -139,7 +139,7 @@ void NVDRV::Close(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -170,7 +170,7 @@ void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 

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

@@ -132,7 +132,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
         multisample = v.X(meta_reg++);
     }
     if (tld.clamp != 0) {
-        throw NotImplementedException("TLD.CL - CLAMP is not implmented");
+        throw NotImplementedException("TLD.CL - CLAMP is not implemented");
     }
     IR::TextureInstInfo info{};
     info.type.Assign(GetType(tld.type));

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

@@ -81,7 +81,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
     } const tmml{insn};
 
     if ((tmml.mask & 0b1100) != 0) {
-        throw NotImplementedException("TMML BA results are not implmented");
+        throw NotImplementedException("TMML BA results are not implemented");
     }
     const IR::Value coords{MakeCoords(v, tmml.coord_reg, tmml.type)};
 

src/video_core/vulkan_common/vulkan_device.cpp

@@ -736,7 +736,7 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags
 }
 
 void Device::ReportLoss() const {
-    LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
+    LOG_CRITICAL(Render_Vulkan, "Device loss occurred!");
 
     // Wait for the log to flush and for Nsight Aftermath to dump the results
     std::this_thread::sleep_for(std::chrono::seconds{15});

src/yuzu/configuration/configure_hotkeys.cpp

@@ -35,8 +35,9 @@ ConfigureHotkeys::ConfigureHotkeys(Core::HID::HIDCore& hid_core, QWidget* parent
     ui->hotkey_list->setContextMenuPolicy(Qt::CustomContextMenu);
     ui->hotkey_list->setModel(model);
 
-    ui->hotkey_list->setColumnWidth(name_column, 200);
+    ui->hotkey_list->header()->setStretchLastSection(false);
-    ui->hotkey_list->resizeColumnToContents(hotkey_column);
+    ui->hotkey_list->header()->setSectionResizeMode(name_column, QHeaderView::ResizeMode::Stretch);
+    ui->hotkey_list->header()->setMinimumSectionSize(150);
 
     connect(ui->button_restore_defaults, &QPushButton::clicked, this,
             &ConfigureHotkeys::RestoreDefaults);
@@ -76,8 +77,8 @@ void ConfigureHotkeys::Populate(const HotkeyRegistry& registry) {
     }
 
     ui->hotkey_list->expandAll();
-    ui->hotkey_list->resizeColumnToContents(name_column);
     ui->hotkey_list->resizeColumnToContents(hotkey_column);
+    ui->hotkey_list->resizeColumnToContents(controller_column);
 }
 
 void ConfigureHotkeys::changeEvent(QEvent* event) {

src/yuzu/configuration/configure_per_game_addons.cpp

@@ -47,6 +47,10 @@ ConfigurePerGameAddons::ConfigurePerGameAddons(Core::System& system_, QWidget* p
     item_model->setHeaderData(0, Qt::Horizontal, tr("Patch Name"));
     item_model->setHeaderData(1, Qt::Horizontal, tr("Version"));
 
+    tree_view->header()->setStretchLastSection(false);
+    tree_view->header()->setSectionResizeMode(0, QHeaderView::ResizeMode::Stretch);
+    tree_view->header()->setMinimumSectionSize(150);
+
     // We must register all custom types with the Qt Automoc system so that we are able to use it
     // with signals/slots. In this case, QList falls under the umbrella of custom types.
     qRegisterMetaType<QList<QStandardItem*>>("QList<QStandardItem*>");
@@ -138,5 +142,5 @@ void ConfigurePerGameAddons::LoadConfiguration() {
         item_model->appendRow(list_items.back());
     }
 
-    tree_view->setColumnWidth(0, 5 * tree_view->width() / 16);
+    tree_view->resizeColumnToContents(1);
 }