fixes for dynarmic on higher batch sizes

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

@@ -6,6 +6,8 @@
  * SPDX-License-Identifier: 0BSD
  */
 
+#pragma once
+
 #include <bit>
 #include "dynarmic/backend/x64/xbyak.h"
 

src/dynarmic/src/dynarmic/common/fp/fused.cpp

@@ -32,7 +32,7 @@ FPUnpacked FusedMulAdd(FPUnpacked addend, FPUnpacked op1, FPUnpacked op2) {
         return std::make_tuple(exponent, value);
     }();
 
-    if (product_value == 0) {
+    if (product_value == u128(0, 0)) {
         return addend;
     }
 
@@ -52,13 +52,13 @@ FPUnpacked FusedMulAdd(FPUnpacked addend, FPUnpacked op1, FPUnpacked op2) {
         }
 
         // addend < product
-        const u128 result = product_value + StickyLogicalShiftRight(addend.mantissa, exp_diff - normalized_point_position);
+        const u128 result = product_value + StickyLogicalShiftRight(u128(addend.mantissa, 0), exp_diff - normalized_point_position);
         return ReduceMantissa(product_sign, product_exponent, result);
     }
 
     // Subtraction
 
-    const u128 addend_long = u128(addend.mantissa) << normalized_point_position;
+    const u128 addend_long = u128(addend.mantissa, 0) << normalized_point_position;
 
     bool result_sign;
     u128 result;

src/dynarmic/src/dynarmic/common/u128.cpp

@@ -137,7 +137,7 @@ u128 StickyLogicalShiftRight(u128 operand, int amount) {
     }
 
     if (operand.lower != 0 || operand.upper != 0) {
-        return u128(1);
+        return u128(1, 0);
     }
     return {};
 }

src/dynarmic/src/dynarmic/common/u128.h

@@ -23,22 +23,13 @@ struct u128 {
     u128(u128&&) = default;
     u128& operator=(const u128&) = default;
     u128& operator=(u128&&) = default;
-
-    u128(u64 lower_, u64 upper_)
-            : lower(lower_), upper(upper_) {}
-
-    template<typename T>
-    /* implicit */ u128(T value)
-            : lower(value), upper(0) {
-        static_assert(std::is_integral_v<T>);
-        static_assert(mcl::bitsizeof<T> <= mcl::bitsizeof<u64>);
-    }
+    explicit u128(u64 lower_, u64 upper_) : lower(lower_), upper(upper_) {}
 
     u64 lower = 0;
     u64 upper = 0;
 
     template<size_t bit_position>
-    bool Bit() const {
+    [[nodiscard]] inline bool Bit() const {
         static_assert(bit_position < 128);
         if constexpr (bit_position < 64) {
             return mcl::bit::get_bit<bit_position>(lower);