early-access version 1465

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 1464.
+This is the source code for early-access 1465.
 
 ## Legal Notice
 

src/common/CMakeLists.txt

@@ -168,7 +168,6 @@ add_library(common STATIC
     time_zone.cpp
     time_zone.h
     tree.h
-    uint128.cpp
     uint128.h
     uuid.cpp
     uuid.h

src/common/uint128.h

@@ -4,12 +4,98 @@
 
 #pragma once
 
+#include <cstring>
 #include <utility>
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#pragma intrinsic(__umulh)
+#pragma intrinsic(_umul128)
+#pragma intrinsic(_udiv128)
+#else
+#include <x86intrin.h>
+#endif
+
 #include "common/common_types.h"
 
 namespace Common {
 
 // This function multiplies 2 u64 values and divides it by a u64 value.
-[[nodiscard]] u64 MultiplyAndDivide64(u64 a, u64 b, u64 d);
+[[nodiscard]] static inline u64 MultiplyAndDivide64(u64 a, u64 b, u64 d) {
+#ifdef _MSC_VER
+    u128 r{};
+    r[0] = _umul128(a, b, &r[1]);
+    u64 remainder;
+#if _MSC_VER < 1923
+    return udiv128(r[1], r[0], d, &remainder);
+#else
+    return _udiv128(r[1], r[0], d, &remainder);
+#endif
+#else
+    const u64 diva = a / d;
+    const u64 moda = a % d;
+    const u64 divb = b / d;
+    const u64 modb = b % d;
+    return diva * b + moda * divb + moda * modb / d;
+#endif
+}
+
+// This function multiplies 2 u64 values and produces a u128 value;
+[[nodiscard]] static inline u128 Multiply64Into128(u64 a, u64 b) {
+    u128 result;
+#ifdef _MSC_VER
+    result[0] = _umul128(a, b, &result[1]);
+#else
+    unsigned __int128 tmp = a;
+    tmp *= b;
+    std::memcpy(&result, &tmp, sizeof(u128));
+#endif
+    return result;
+}
+
+[[nodiscard]] static inline u64 GetFixedPoint64Factor(u64 numerator, u64 divisor) {
+#ifdef __SIZEOF_INT128__
+    const auto base = static_cast<unsigned __int128>(numerator) << 64ULL;
+    return static_cast<u64>(base / divisor);
+#elif defined(_M_X64) || defined(_M_ARM64)
+    std::array<u64, 2> r = {0, numerator};
+    u64 remainder;
+#if _MSC_VER < 1923
+    return udiv128(r[1], r[0], divisor, &remainder);
+#else
+    return _udiv128(r[1], r[0], divisor, &remainder);
+#endif
+#else
+    // This one is bit more inaccurate.
+    return MultiplyAndDivide64(std::numeric_limits<u64>::max(), numerator, divisor);
+#endif
+}
+
+[[nodiscard]] static inline u64 MultiplyHigh(u64 a, u64 b) {
+#ifdef __SIZEOF_INT128__
+    return (static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b)) >> 64;
+#elif defined(_M_X64) || defined(_M_ARM64)
+    return __umulh(a, b); // MSVC
+#else
+    // Generic fallback
+    const u64 a_lo = u32(a);
+    const u64 a_hi = a >> 32;
+    const u64 b_lo = u32(b);
+    const u64 b_hi = b >> 32;
+
+    const u64 a_x_b_hi = a_hi * b_hi;
+    const u64 a_x_b_mid = a_hi * b_lo;
+    const u64 b_x_a_mid = b_hi * a_lo;
+    const u64 a_x_b_lo = a_lo * b_lo;
+
+    const u64 carry_bit = (static_cast<u64>(static_cast<u32>(a_x_b_mid)) +
+                           static_cast<u64>(static_cast<u32>(b_x_a_mid)) + (a_x_b_lo >> 32)) >>
+                          32;
+
+    const u64 multhi = a_x_b_hi + (a_x_b_mid >> 32) + (b_x_a_mid >> 32) + carry_bit;
+
+    return multhi;
+#endif
+}
 
 } // namespace Common

src/common/wall_clock.cpp

@@ -2,6 +2,9 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cstdint>
+
+#include "common/uint128.h"
 #include "common/wall_clock.h"
 
 #ifdef ARCHITECTURE_x86_64
@@ -17,7 +20,9 @@ using base_time_point = std::chrono::time_point<base_timer>;
 class StandardWallClock final : public WallClock {
 public:
     explicit StandardWallClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequency_)
-        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false) {
+        : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, false),
+          emulated_clock_factor{GetFixedPoint64Factor(emulated_clock_frequency, 1000000000)},
+          emulated_cpu_factor{GetFixedPoint64Factor(emulated_cpu_frequency, 1000000000)} {
         start_time = base_timer::now();
     }
 
@@ -40,11 +45,11 @@ public:
     }
 
     u64 GetClockCycles() override {
-        return GetTimeNS().count() * (emulated_clock_frequency / 1000) / 1000000;
+        return MultiplyHigh(GetTimeNS().count(), emulated_clock_factor);
     }
 
     u64 GetCPUCycles() override {
-        return GetTimeNS().count() * (emulated_cpu_frequency / 1000) / 1000000;
+        return MultiplyHigh(GetTimeNS().count(), emulated_cpu_factor);
     }
 
     void Pause([[maybe_unused]] bool is_paused) override {
@@ -53,6 +58,8 @@ public:
 
 private:
     base_time_point start_time;
+    const u64 emulated_clock_factor;
+    const u64 emulated_cpu_factor;
 };
 
 #ifdef ARCHITECTURE_x86_64

src/common/x64/native_clock.cpp

@@ -8,68 +8,10 @@
 #include <mutex>
 #include <thread>
 
-#ifdef _MSC_VER
-#include <intrin.h>
-
-#pragma intrinsic(__umulh)
-#pragma intrinsic(_udiv128)
-#else
-#include <x86intrin.h>
-#endif
-
 #include "common/atomic_ops.h"
 #include "common/uint128.h"
 #include "common/x64/native_clock.h"
 
-namespace {
-
-[[nodiscard]] u64 GetFixedPoint64Factor(u64 numerator, u64 divisor) {
-#ifdef __SIZEOF_INT128__
-    const auto base = static_cast<unsigned __int128>(numerator) << 64ULL;
-    return static_cast<u64>(base / divisor);
-#elif defined(_M_X64) || defined(_M_ARM64)
-    std::array<u64, 2> r = {0, numerator};
-    u64 remainder;
-#if _MSC_VER < 1923
-    return udiv128(r[1], r[0], divisor, &remainder);
-#else
-    return _udiv128(r[1], r[0], divisor, &remainder);
-#endif
-#else
-    // This one is bit more inaccurate.
-    return MultiplyAndDivide64(std::numeric_limits<u64>::max(), numerator, divisor);
-#endif
-}
-
-[[nodiscard]] u64 MultiplyHigh(u64 a, u64 b) {
-#ifdef __SIZEOF_INT128__
-    return (static_cast<unsigned __int128>(a) * static_cast<unsigned __int128>(b)) >> 64;
-#elif defined(_M_X64) || defined(_M_ARM64)
-    return __umulh(a, b); // MSVC
-#else
-    // Generic fallback
-    const u64 a_lo = u32(a);
-    const u64 a_hi = a >> 32;
-    const u64 b_lo = u32(b);
-    const u64 b_hi = b >> 32;
-
-    const u64 a_x_b_hi = a_hi * b_hi;
-    const u64 a_x_b_mid = a_hi * b_lo;
-    const u64 b_x_a_mid = b_hi * a_lo;
-    const u64 a_x_b_lo = a_lo * b_lo;
-
-    const u64 carry_bit = (static_cast<u64>(static_cast<u32>(a_x_b_mid)) +
-                           static_cast<u64>(static_cast<u32>(b_x_a_mid)) + (a_x_b_lo >> 32)) >>
-                          32;
-
-    const u64 multhi = a_x_b_hi + (a_x_b_mid >> 32) + (b_x_a_mid >> 32) + carry_bit;
-
-    return multhi;
-#endif
-}
-
-} // namespace
-
 namespace Common {
 
 u64 EstimateRDTSCFrequency() {