nx2: run the guest counter at 31.25 MHz for Switch 2 titles

The Switch 2 counter runs at 31.25 MHz and its SDK converts ticks with
that constant: Mario Kart World's sdk builds 31,250,000 in
nn::os::GetSystemTickFrequency, TickManager::ConvertToTimeSpan and
TickManager::ConvertToTick (19.2 MHz appears nowhere). Eglantine counted
at the Switch 1 rate for every title, so every tick-based time in a
Switch 2 game was off by 1.63x: elapsed times read short, and a deadline
of one 60 Hz frame converted to ticks waited ~27 ms.

The wall clock now has a guest counter frequency, set at boot: 31.25 MHz
for Switch 2 programs, 19.2 MHz otherwise, EDEN_NX2_CNTFRQ=<Hz> to force
one. CNTPCT reads, CNTFRQ_EL0, svcGetSystemTick, thread tick counts and
the time services' tick conversions follow it, and the CNTPCT spin
detection keeps its ~1.7 us threshold.

Mario Kart World, title screen and attract demo, 180 s, same build:
frame times unchanged (17-19 ms once warm), process CPU time 848 s
instead of 930 s.

6 files changed, 36 insertions, 12 deletions

diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
index a2c0124f33..e65e38dd9e 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_64.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
@@ -180,7 +180,7 @@ u64 DynarmicCallbacks64::GetCNTPCT() {
         const char* value = std::getenv("EDEN_NX2_CNTPCT_PAUSE");
         return value == nullptr || value[0] != '0';
     }();
-    if (pause_spins && ticks - m_last_cntpct < 32) {
+    if (pause_spins && ticks - m_last_cntpct < m_spin_ticks) {
 #if defined(_M_X64) || defined(__x86_64__)
         _mm_pause();
         _mm_pause();
@@ -269,7 +269,9 @@ void ArmDynarmic64::MakeJit(Common::PageTable* page_table, std::size_t address_s
     config.tpidr_el0 = &m_cb->m_tpidr_el0;
     config.dczid_el0 = 4;
     config.ctr_el0 = 0x8444c004;
-    config.cntfrq_el0 = Hardware::CNTFREQ;
+    // 31.25 MHz for Switch 2 titles (Common::WallClock::SetGuestCNTFRQ, set at boot).
+    config.cntfrq_el0 = static_cast<u32>(Common::g_wall_clock.GetGuestCNTFRQ());
+    m_cb->m_spin_ticks = config.cntfrq_el0 / 600'000;
 
     // Unpredictable instructions
     config.define_unpredictable_behaviour = true;
diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.h b/src/core/arm/dynarmic/arm_dynarmic_64.h
index acd1fe773f..66b07474c1 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_64.h
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.h
@@ -66,6 +66,7 @@ public:
     u64 m_tpidrro_el0{};
     u64 m_tpidr_el0{};
     u64 m_last_cntpct{}; ///< The previous counter read (GetCNTPCT spin detection)
+    u64 m_spin_ticks{32}; ///< Reads closer than this (~1.7 us) count as a polling loop
     Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
diff --git a/src/core/arm/nce/patcher.cpp b/src/core/arm/nce/patcher.cpp
index f21d7cacdc..72c3497f03 100644
--- a/src/core/arm/nce/patcher.cpp
+++ b/src/core/arm/nce/patcher.cpp
@@ -627,7 +627,7 @@ void Patcher::WriteMsrHandler(ModuleDestLabel module_dest, oaknut::XReg src_reg,
 }
 
 void Patcher::WriteCntfrqHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::VectorCodeGenerator& cg) {
-    cg.MOV(dest_reg, Common::WallClock::CNTFRQ);
+    cg.MOV(dest_reg, Common::g_wall_clock.GetGuestCNTFRQ());
 
     // Jump back to the instruction after the emulated MRS.
     if (&cg == &c_pre)
@@ -642,6 +642,7 @@ void Patcher::WriteCntpctHandler(ModuleDestLabel module_dest, oaknut::XReg dest_
 #else
     static Common::WallClock clock(true, 1);
 #endif
+    clock.SetGuestCNTFRQ(Common::g_wall_clock.GetGuestCNTFRQ());
     const auto factor = clock.GetGuestCNTFRQFactor();
     const auto raw_factor = std::bit_cast<std::array<u64, 2>>(factor);
 
diff --git a/src/core/core.cpp b/src/core/core.cpp
index f3d224b1ce..3728ccf8f7 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -2,11 +2,13 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 #include <array>
+#include <cstdlib>
 #include <atomic>
 #include <memory>
 #include <utility>
 
 #include "game_settings.h"
+#include "common/cpu_features.h"
 #include "audio_core/audio_core.h"
 #include "common/fs/fs.h"
 #include "common/logging.h"
@@ -313,16 +315,33 @@ struct System::Impl {
             fs_controller.InitTempStorage();
         }
 
+        const bool switch2_program =
+            virtual_filesystem != nullptr &&
+            IsSwitch2Program(GetGameFileFromPath(virtual_filesystem, filepath));
+
         // Switch 2 titles size their heap from the total memory and need the 12 GiB layout. Set it
         // as a per-game value, so RestoreGlobalState brings back the configured layout afterwards.
-        if (virtual_filesystem != nullptr &&
-            Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_12Gb &&
-            IsSwitch2Program(GetGameFileFromPath(virtual_filesystem, filepath))) {
+        if (switch2_program &&
+            Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_12Gb) {
             LOG_INFO(Core, "Switch 2 program: using the 12 GiB memory layout");
             Settings::values.memory_layout_mode.SetGlobal(false);
             Settings::values.memory_layout_mode.SetValue(Settings::MemoryLayout::Memory_12Gb);
         }
 
+        // The Switch 2 counter runs at 31.25 MHz and its SDK converts ticks with that constant
+        // (nn::os::GetSystemTickFrequency, TickManager::ConvertToTimeSpan/ConvertToTick). At 19.2 MHz
+        // every tick-based time in the game was off by 1.63x: elapsed times read short and a frame
+        // deadline in ticks waited ~27 ms. EDEN_NX2_CNTFRQ=<Hz> overrides it (A/B tests).
+        static const u64 forced_cntfrq = [] {
+            const char* value = std::getenv("EDEN_NX2_CNTFRQ");
+            return value != nullptr ? std::strtoull(value, nullptr, 10) : 0ULL;
+        }();
+        const u64 cntfrq = forced_cntfrq != 0 ? forced_cntfrq
+                           : switch2_program  ? Common::WallClock::Switch2CNTFRQ
+                                              : Common::WallClock::CNTFRQ;
+        Common::g_wall_clock.SetGuestCNTFRQ(cntfrq);
+        LOG_INFO(Core, "Guest counter frequency: {} Hz", cntfrq);
+
         InitializeKernel(system);
 
         if (params.applet_type == Service::AM::AppletType::Application) {
diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp
index b85bd30a58..dfc3f6f2a6 100644
--- a/src/core/core_timing.cpp
+++ b/src/core/core_timing.cpp
@@ -224,7 +224,7 @@ void CoreTiming::ResetTicks() {
 }
 
 u64 CoreTiming::GetClockTicks() const {
-    u64 fres = is_multicore ? Common::g_wall_clock.GetCNTPCT() : Common::WallClock::CPUTickToCNTPCT(cpu_ticks);
+    u64 fres = is_multicore ? Common::g_wall_clock.GetCNTPCT() : Common::g_wall_clock.CPUTickToGuestCNTPCT(cpu_ticks);
     if (const u32 target = CpuClockTargetMhz(Settings::values.cpu_clock.GetValue());
         target != CPU_CLOCK_BASE_MHZ) {
         fres = fres * target / CPU_CLOCK_BASE_MHZ;
diff --git a/src/core/hle/service/psc/time/common.h b/src/core/hle/service/psc/time/common.h
index 2183fa72b8..6f7d2cb88b 100644
--- a/src/core/hle/service/psc/time/common.h
+++ b/src/core/hle/service/psc/time/common.h
@@ -134,12 +134,13 @@ struct OperationEvent : public Common::IntrusiveListBaseNode<OperationEvent> {
     Kernel::KEvent* m_event{};
 };
 
-constexpr inline std::chrono::nanoseconds ConvertToTimeSpan(s64 ticks) {
+inline std::chrono::nanoseconds ConvertToTimeSpan(s64 ticks) {
     constexpr auto one_second_ns{
         std::chrono::duration_cast<std::chrono::nanoseconds>(std::chrono::seconds(1)).count()};
 
-    constexpr s64 max{Common::WallClock::CNTFRQ *
-                      ((std::numeric_limits<s64>::max)() / one_second_ns)};
+    // The guest counter's frequency: 31.25 MHz for Switch 2 titles.
+    const s64 frequency{static_cast<s64>(Common::g_wall_clock.GetGuestCNTFRQ())};
+    const s64 max{frequency * ((std::numeric_limits<s64>::max)() / one_second_ns)};
 
     if (ticks > max) {
         return std::chrono::nanoseconds((std::numeric_limits<s64>::max)());
@@ -147,8 +148,8 @@ constexpr inline std::chrono::nanoseconds ConvertToTimeSpan(s64 ticks) {
         return std::chrono::nanoseconds((std::numeric_limits<s64>::min)());
     }
 
-    auto a{ticks / Common::WallClock::CNTFRQ * one_second_ns};
-    auto b{((ticks % Common::WallClock::CNTFRQ) * one_second_ns) / Common::WallClock::CNTFRQ};
+    auto a{ticks / frequency * one_second_ns};
+    auto b{((ticks % frequency) * one_second_ns) / frequency};
 
     return std::chrono::nanoseconds(a + b);
 }