Merge branch 'master' into headers

BUILDING.md

@@ -8,11 +8,12 @@ Other instructions may be found [here](https://wiki.rpcs3.net/index.php?title=Bu
 ### Windows 10 or later
 
 The following tools are required to build RPCS3 on Windows 10 or later:
-- [Visual Studio 2022](https://visualstudio.microsoft.com/thank-you-downloading-visual-studio/?sku=Community)
+- [Visual Studio 2022/2026](https://visualstudio.microsoft.com/thank-you-downloading-visual-studio/?sku=Community)
 - **Optional** - [CMake 3.28.0+](https://www.cmake.org/download/) (add to PATH)
 
   **NOTES:**
-  - **Visual Studio 2022** integrates **CMake 3.29+** and it also supports both the `sln` solution (`.sln`, `.vcxproj`) and `CMake` solution (`CMakeLists.txt`, `CMakePresets.json`).
+  - **Visual Studio 2026** needs at least **CMake 4.2.0+**.
+  - **Visual Studio 2022/2026** integrates **CMake 3.29+** and it also supports both the `sln` solution (`.sln`, `.vcxproj`) and `CMake` solution (`CMakeLists.txt`, `CMakePresets.json`).
     See sections [Building with Visual Studio sln solution](#building-with-visual-studio-sln-solution) and [Building with Visual Studio CMake solution](#building-with-visual-studio-cmake-solution)
     on how to build the project with **Visual Studio**.
   - Install and use this standalone **CMake** tool just in case of your preference. See section [Building with standalone CMake tool](#building-with-standalone-cmake-tool) on how to build the project

rpcs3/Emu/Cell/SPUCommonRecompiler.cpp

@@ -720,9 +720,19 @@ void spu_cache::initialize(bool build_existing_cache)
 	}
 
 	// SPU cache file (version + block size type)
-	const std::string loc = ppu_cache + "spu-" + fmt::to_lower(g_cfg.core.spu_block_size.to_string()) + "-v1-tane.dat";
+	const std::string filename = "spu-" + fmt::to_lower(g_cfg.core.spu_block_size.to_string()) + "-v1-tane.dat";
+	const std::string loc = ppu_cache + filename;
+	const std::string loc_debug = fs::get_cache_dir() + "DEBUG/" + filename;
 
-	spu_cache cache(loc);
+	bool is_debug = false;
+
+	if (fs::is_file(loc_debug))
+	{
+		spu_log.success("SPU Cache override applied!");
+		is_debug = true;
+	}
+
+	spu_cache cache(is_debug ? loc_debug : loc);
 
 	if (!cache)
 	{
@@ -4963,6 +4973,7 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 		u32 lsa_last_pc = SPU_LS_SIZE; // PC of first LSA write
 		u32 get_pc = SPU_LS_SIZE; // PC of GETLLAR
 		u32 put_pc = SPU_LS_SIZE; // PC of PUTLLC
+		u32 rdatomic_pc = SPU_LS_SIZE; // PC of last RdAtomcStat read
 		reg_state_t ls{}; // state of LS load/store address register
 		reg_state_t ls_offs = reg_state_t::from_value(0); // Added value to ls
 		reg_state_t lsa{}; // state of LSA register on GETLLAR
@@ -5008,7 +5019,7 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 			ls_invalid = true;
 			ls_write |= write;
 
-			if (write)
+			if (ls_write)
 			{
 				return discard();
 			}
@@ -6323,6 +6334,8 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 							break;
 						}
 
+						atomic16->rdatomic_pc = pos;
+
 						const auto it = atomic16_all.find(pos);
 
 						if (it == atomic16_all.end())
@@ -7263,7 +7276,7 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 
 	for (const auto& [pc_commited, pattern] : atomic16_all)
 	{
-		if (!pattern.active)
+		if (!pattern.active || pattern.lsa_pc >= pattern.rdatomic_pc)
 		{
 			continue;
 		}
@@ -7273,6 +7286,17 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 			continue;
 		}
 
+		std::string pattern_hash;
+		{
+			sha1_context ctx;
+			u8 output[20]{};
+
+			sha1_starts(&ctx);
+			sha1_update(&ctx, reinterpret_cast<const u8*>(result.data.data()) + (pattern.lsa_pc - result.lower_bound), pattern.rdatomic_pc - pattern.lsa_pc);
+			sha1_finish(&ctx, output);
+			fmt::append(pattern_hash, "%s", fmt::base57(output));
+		}
+
 		union putllc16_or_0_info
 		{
 			u64 data;
@@ -7295,8 +7319,8 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 				value.required_pc = pattern.required_pc;
 			}
 
-			spu_log.success("PUTLLC0 Pattern Detected! (put_pc=0x%x, %s) (putllc0=%d, putllc16+0=%d, all=%d)", pattern.put_pc, func_hash, ++stats.nowrite, ++stats.single, +stats.all);
+			// spu_log.success("PUTLLC0 Pattern Detected! (put_pc=0x%x, %s) (putllc0=%d, putllc16+0=%d, all=%d)", pattern.put_pc, func_hash, ++stats.nowrite, ++stats.single, +stats.all);
-			add_pattern(false, inst_attr::putllc0, pattern.put_pc - lsa, value.data);
+			// add_pattern(false, inst_attr::putllc0, pattern.put_pc - lsa, value.data);
 			continue;
 		}
 
@@ -7363,16 +7387,35 @@ spu_program spu_recompiler_base::analyse(const be_t<u32>* ls, u32 entry_point, s
 			value.reg2 = pattern.reg2;
 		}
 
+		bool allow_pattern = true;
+
 		if (g_cfg.core.spu_accurate_reservations)
 		{
-			// Because enabling it is a hack, as it turns out
+			// The problem with PUTLLC16 optimization, that it is in theory correct at the bounds of the spu function.
-			// continue;
+			// But if the SPU code reuses the cache line data observed, it is not truly atomic.
+			// So we may enable it only for known cases where SPU atomic data is not used after the function leaves.
+
+			// So the two options are:
+
+			// 1. Atomic compare exchange 16 bytes operation. (rest of data is not read) -> good for RPCS3 to optimize.
+			// 2. Fetch 128 bytes (read them later), modify only 16 bytes. -> Bad for RPCS3 to optimize.
+
+			// This difference cannot be known at analyzer time but from observing callers.
+			static constexpr std::initializer_list<std::string_view> allowed_patterns =
+			{
+				"620oYSe8uQqq9eTkhWfMqoEXX0us"sv, // CellSpurs JobChain acquire pattern
+			};
+
+			allow_pattern = std::any_of(allowed_patterns.begin(), allowed_patterns.end(), FN(pattern_hash == x));
 		}
 
-		add_pattern(false, inst_attr::putllc16, pattern.put_pc - result.entry_point, value.data);
+		if (allow_pattern)
+		{
+			add_pattern(false, inst_attr::putllc16, pattern.put_pc - result.entry_point, value.data);
+		}
 
-		spu_log.success("PUTLLC16 Pattern Detected! (mem_count=%d, put_pc=0x%x, pc_rel=%d, offset=0x%x, const=%u, two_regs=%d, reg=%u, runtime=%d, 0x%x-%s) (putllc0=%d, putllc16+0=%d, all=%d)"
+		spu_log.success("PUTLLC16 Pattern Detected! (mem_count=%d, put_pc=0x%x, pc_rel=%d, offset=0x%x, const=%u, two_regs=%d, reg=%u, runtime=%d, 0x%x-%s, pattern-hash=%s) (putllc0=%d, putllc16+0=%d, all=%d)"
-			, pattern.mem_count, pattern.put_pc, value.type == v_relative, value.off18, value.type == v_const, value.type == v_reg2, value.reg, value.runtime16_select, entry_point, func_hash, +stats.nowrite, ++stats.single, +stats.all);
+			, pattern.mem_count, pattern.put_pc, value.type == v_relative, value.off18, value.type == v_const, value.type == v_reg2, value.reg, value.runtime16_select, entry_point, func_hash, pattern_hash, +stats.nowrite, ++stats.single, +stats.all);
 	}
 
 	for (const auto& [read_pc, pattern] : rchcnt_loop_all)

rpcs3/rpcs3qt/emu_settings.cpp

@@ -668,10 +668,9 @@ void emu_settings::EnhanceSpinBox(QSpinBox* spinbox, emu_settings_type type, con
 	spinbox->setRange(min, max);
 	spinbox->setValue(val);
 
-	connect(spinbox, &QSpinBox::textChanged, this, [type, spinbox, this](const QString& /* text*/)
+	connect(spinbox, &QSpinBox::valueChanged, this, [type, this](int value)
 	{
-		if (!spinbox) return;
+		SetSetting(type, fmt::format("%d", value));
-		SetSetting(type, spinbox->cleanText().toStdString());
 	});
 
 	connect(this, &emu_settings::RestoreDefaultsSignal, spinbox, [def, spinbox]()
@@ -724,10 +723,9 @@ void emu_settings::EnhanceDoubleSpinBox(QDoubleSpinBox* spinbox, emu_settings_ty
 	spinbox->setRange(min, max);
 	spinbox->setValue(val);
 
-	connect(spinbox, &QDoubleSpinBox::textChanged, this, [type, spinbox, this](const QString& /* text*/)
+	connect(spinbox, &QDoubleSpinBox::valueChanged, this, [type, this](double value)
 	{
-		if (!spinbox) return;
+		SetSetting(type, fmt::format("%f", value));
-		SetSetting(type, spinbox->cleanText().toStdString());
 	});
 
 	connect(this, &emu_settings::RestoreDefaultsSignal, spinbox, [def, spinbox]()

rpcs3/rpcs3qt/memory_viewer_panel.cpp

@@ -432,9 +432,9 @@ memory_viewer_panel::memory_viewer_panel(QWidget* parent, std::shared_ptr<CPUDis
 
 		scroll(0); // Refresh
 	});
-	connect(sb_words, static_cast<void (QSpinBox::*)(int)>(&QSpinBox::valueChanged), this, [=, this]()
+	connect(sb_words, &QSpinBox::valueChanged, this, [this](int value)
 	{
-		m_colcount = 1 << sb_words->value();
+		m_colcount = 1 << value;
 		ShowMemory();
 	});
 

rpcs3/rpcs3qt/pad_motion_settings_dialog.cpp

@@ -105,7 +105,7 @@ pad_motion_settings_dialog::pad_motion_settings_dialog(QDialog* parent, std::sha
 				m_config_entries[i]->mirrored.set(state != Qt::Unchecked);
 			});
 
-			connect(m_shifts[i], QOverload<int>::of(&QSpinBox::valueChanged), this, [this, i](int value)
+			connect(m_shifts[i], &QSpinBox::valueChanged, this, [this, i](int value)
 			{
 				std::lock_guard lock(m_config_mutex);
 				m_config_entries[i]->shift.set(value);

rpcs3/rpcs3qt/patch_manager_dialog.cpp

@@ -123,8 +123,8 @@ patch_manager_dialog::patch_manager_dialog(std::shared_ptr<gui_settings> gui_set
 			handle_config_value_changed(ui->configurable_combo_box->itemData(index).toDouble());
 		}
 	});
-	connect(ui->configurable_spin_box, QOverload<int>::of(&QSpinBox::valueChanged), this, &patch_manager_dialog::handle_config_value_changed);
+	connect(ui->configurable_spin_box, &QSpinBox::valueChanged, this, &patch_manager_dialog::handle_config_value_changed);
-	connect(ui->configurable_double_spin_box, QOverload<double>::of(&QDoubleSpinBox::valueChanged), this, &patch_manager_dialog::handle_config_value_changed);
+	connect(ui->configurable_double_spin_box, &QDoubleSpinBox::valueChanged, this, &patch_manager_dialog::handle_config_value_changed);
 	connect(ui->buttonBox, &QDialogButtonBox::rejected, this, &QWidget::close);
 	connect(ui->buttonBox, &QDialogButtonBox::clicked, [this](QAbstractButton* button)
 	{

rpcs3/rpcs3qt/raw_mouse_settings_dialog.cpp

@@ -281,7 +281,7 @@ void raw_mouse_settings_dialog::add_tabs(QTabWidget* tabs)
 		m_accel_spin_boxes.push_back(mouse_acceleration_spin_box);
 		mouse_acceleration_spin_box->setRange(0.1, 10.0);
 		mouse_acceleration_spin_box->setValue(config->mouse_acceleration.get() / 100.0);
-		connect(mouse_acceleration_spin_box, QOverload<double>::of(&QDoubleSpinBox::valueChanged), this, [player](double value)
+		connect(mouse_acceleration_spin_box, &QDoubleSpinBox::valueChanged, this, [player](double value)
 		{
 			auto& config = ::at32(g_cfg_raw_mouse.players, player)->mouse_acceleration;
 			config.set(std::clamp(value * 100.0, config.min, config.max));