diff --git a/stl/inc/random b/stl/inc/random
index f866dab8eaa..90ae36415a5 100644
--- a/stl/inc/random
+++ b/stl/inc/random
@@ -461,14 +461,6 @@ _NODISCARD _Real _Nrand_impl(_Gen& _Gx) { // build a floating-point value from r
     }
 }
 
-_INLINE_VAR constexpr int _MP_len = 5;
-using _MP_arr                     = uint64_t[_MP_len];
-
-extern "C++" _NODISCARD _CRTIMP2_PURE uint64_t __CLRCALL_PURE_OR_CDECL _MP_Get(_MP_arr) noexcept;
-extern "C++" _CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Add(_MP_arr, uint64_t) noexcept;
-extern "C++" _CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Mul(_MP_arr, uint64_t, uint64_t) noexcept;
-extern "C++" _CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Rem(_MP_arr, uint64_t) noexcept;
-
 template <class _Uint, _Uint _Ax, _Uint _Cx, _Uint _Mx>
 _NODISCARD _Uint _Next_linear_congruential_value(_Uint _Prev) noexcept {
     // Choose intermediate type:
@@ -500,16 +492,13 @@ _NODISCARD _Uint _Next_linear_congruential_value(_Uint _Prev) noexcept {
         const auto _Mul =
             static_cast<unsigned int>(_Prev) * static_cast<unsigned int>(_Ax) + static_cast<unsigned int>(_Cx);
         return static_cast<_Uint>(_Mul % _Mx);
-    } else if constexpr (_Cx <= ULLONG_MAX && static_cast<_Uint>(_Mx - 1) <= (ULLONG_MAX - _Cx) / _Ax) {
+    } else if constexpr (static_cast<_Uint>(_Mx - 1) <= (ULLONG_MAX - _Cx) / _Ax) {
         // unsigned long long is sufficient to store intermediate calculation
         const auto _Mul = static_cast<unsigned long long>(_Prev) * _Ax + _Cx;
         return static_cast<_Uint>(_Mul % _Mx);
     } else { // no intermediate integral type fits; fall back to multiprecision
-        _MP_arr _Wx;
-        _MP_Mul(_Wx, _Prev, _Ax);
-        _MP_Add(_Wx, _Cx);
-        _MP_Rem(_Wx, _Mx);
-        return static_cast<_Uint>(_MP_Get(_Wx));
+        const auto _Mul = _Unsigned128{_Prev} * _Ax + _Cx;
+        return static_cast<_Uint>(_Mul % _Mx);
     }
 }
 
diff --git a/stl/src/multprec.cpp b/stl/src/multprec.cpp
index 0043cfbfb96..326fd9cbb7c 100644
--- a/stl/src/multprec.cpp
+++ b/stl/src/multprec.cpp
@@ -3,15 +3,18 @@
 
 // implements multiprecision math for random number generators
 
-#include <limits>
-#include <random>
+#include <yvals.h>
 
 _STD_BEGIN
-constexpr int shift                 = _STD numeric_limits<unsigned long long>::digits / 2;
+constexpr int _MP_len = 5;
+using _MP_arr         = unsigned long long[_MP_len];
+
+constexpr int shift                 = 64 / 2;
 constexpr unsigned long long mask   = ~(~0ULL << shift);
 constexpr unsigned long long maxVal = mask + 1;
 
-[[nodiscard]] unsigned long long __CLRCALL_PURE_OR_CDECL _MP_Get(
+// TRANSITION, ABI: preserved for binary compatibility
+[[nodiscard]] _CRTIMP2_PURE unsigned long long __CLRCALL_PURE_OR_CDECL _MP_Get(
     _MP_arr u) noexcept { // convert multi-word value to scalar value
     return (u[1] << shift) + u[0];
 }
@@ -32,7 +35,8 @@ static void add(unsigned long long* u, int ulen, unsigned long long* v,
     }
 }
 
-void __CLRCALL_PURE_OR_CDECL _MP_Add(
+// TRANSITION, ABI: preserved for binary compatibility
+_CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Add(
     _MP_arr u, unsigned long long v0) noexcept { // add scalar value to multi-word value
     unsigned long long v[2];
     v[0] = v0 & mask;
@@ -50,7 +54,8 @@ static void mul(
     }
 }
 
-void __CLRCALL_PURE_OR_CDECL _MP_Mul(
+// TRANSITION, ABI: preserved for binary compatibility
+_CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Mul(
     _MP_arr w, unsigned long long u0, unsigned long long v0) noexcept { // multiply multi-word value by multi-word value
     constexpr int m = 2;
     constexpr int n = 2;
@@ -106,7 +111,8 @@ static void div(_MP_arr u,
     return ulen;
 }
 
-void __CLRCALL_PURE_OR_CDECL _MP_Rem(
+// TRANSITION, ABI: preserved for binary compatibility
+_CRTIMP2_PURE void __CLRCALL_PURE_OR_CDECL _MP_Rem(
     _MP_arr u, unsigned long long v0) noexcept { // divide multi-word value by value, leaving remainder in u
     unsigned long long v[2];
     v[0]        = v0 & mask;