diff --git a/java/jni/YGJNIVanilla.cpp b/java/jni/YGJNIVanilla.cpp
index 46d2248a..6a1406c1 100644
--- a/java/jni/YGJNIVanilla.cpp
+++ b/java/jni/YGJNIVanilla.cpp
@@ -16,6 +16,7 @@
 #include "YogaJniException.h"
 
 #include <yoga/Yoga-internal.h>
+#include <yoga/bits/BitCast.h>
 
 // TODO: Reconcile missing layoutContext functionality from callbacks in the C
 // API and use that
@@ -677,11 +678,10 @@ static YGSize YGJNIMeasureFunc(
     uint32_t wBits = 0xFFFFFFFF & (measureResult >> 32);
     uint32_t hBits = 0xFFFFFFFF & measureResult;
 
-    // TODO: this is unsafe under strict aliasing and should use bit_cast
-    const float* measuredWidth = reinterpret_cast<float*>(&wBits);
-    const float* measuredHeight = reinterpret_cast<float*>(&hBits);
+    const float measuredWidth = yoga::bit_cast<float>(wBits);
+    const float measuredHeight = yoga::bit_cast<float>(hBits);
 
-    return YGSize{*measuredWidth, *measuredHeight};
+    return YGSize{measuredWidth, measuredHeight};
   } else {
     return YGSize{
         widthMode == YGMeasureModeUndefined ? 0 : width,
diff --git a/yoga/bits/BitCast.h b/yoga/bits/BitCast.h
new file mode 100644
index 00000000..eaca348c
--- /dev/null
+++ b/yoga/bits/BitCast.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#pragma once
+
+#include <cstring>
+#include <type_traits>
+
+namespace facebook::yoga {
+
+// Polyfill for std::bit_cast() from C++20, to allow safe type punning
+// https://en.cppreference.com/w/cpp/numeric/bit_cast
+template <class To, class From>
+std::enable_if_t<
+    sizeof(To) == sizeof(From) && std::is_trivially_copyable_v<From> &&
+        std::is_trivially_copyable_v<To> &&
+        std::is_trivially_constructible_v<To>,
+    To>
+bit_cast(const From& src) noexcept {
+  To dst;
+  std::memcpy(&dst, &src, sizeof(To));
+  return dst;
+}
+
+} // namespace facebook::yoga
diff --git a/yoga/style/CompactValue.h b/yoga/style/CompactValue.h
index 498b984a..282a518f 100644
--- a/yoga/style/CompactValue.h
+++ b/yoga/style/CompactValue.h
@@ -14,19 +14,7 @@
 #include <yoga/YGMacros.h>
 #include <yoga/YGValue.h>
 
-#if defined(__has_include) && __has_include(<version>)
-// needed to be able to evaluate defined(__cpp_lib_bit_cast)
-#include <version>
-#else
-// needed to be able to evaluate defined(__cpp_lib_bit_cast)
-#include <ciso646>
-#endif
-
-#ifdef __cpp_lib_bit_cast
-#include <bit>
-#else
-#include <cstring>
-#endif
+#include <yoga/bits/BitCast.h>
 
 static_assert(
     std::numeric_limits<float>::is_iec559,
@@ -76,7 +64,7 @@ public:
     }
 
     uint32_t unitBit = Unit == YGUnitPercent ? PERCENT_BIT : 0;
-    auto data = asU32(value);
+    auto data = yoga::bit_cast<uint32_t>(value);
     data -= BIAS;
     data |= unitBit;
     return {data};
@@ -129,7 +117,7 @@ public:
         return YGValue{0.0f, YGUnitPercent};
     }
 
-    if (std::isnan(asFloat(repr_))) {
+    if (std::isnan(yoga::bit_cast<float>(repr_))) {
       return YGValueUndefined;
     }
 
@@ -138,13 +126,14 @@ public:
     data += BIAS;
 
     return YGValue{
-        asFloat(data), repr_ & 0x40000000 ? YGUnitPercent : YGUnitPoint};
+        yoga::bit_cast<float>(data),
+        repr_ & 0x40000000 ? YGUnitPercent : YGUnitPoint};
   }
 
   bool isUndefined() const noexcept {
     return (
         repr_ != AUTO_BITS && repr_ != ZERO_BITS_POINT &&
-        repr_ != ZERO_BITS_PERCENT && std::isnan(asFloat(repr_)));
+        repr_ != ZERO_BITS_PERCENT && std::isnan(yoga::bit_cast<float>(repr_)));
   }
 
   bool isAuto() const noexcept { return repr_ == AUTO_BITS; }
@@ -164,30 +153,6 @@ private:
   constexpr CompactValue(uint32_t data) noexcept : repr_(data) {}
 
   VISIBLE_FOR_TESTING uint32_t repr() { return repr_; }
-
-  static uint32_t asU32(float f) {
-#ifdef __cpp_lib_bit_cast
-    return std::bit_cast<uint32_t>(f);
-#else
-    uint32_t u;
-    static_assert(
-        sizeof(u) == sizeof(f), "uint32_t and float must have the same size");
-    std::memcpy(&u, &f, sizeof(f));
-    return u;
-#endif
-  }
-
-  static float asFloat(uint32_t u) {
-#ifdef __cpp_lib_bit_cast
-    return std::bit_cast<float>(u);
-#else
-    float f;
-    static_assert(
-        sizeof(f) == sizeof(u), "uint32_t and float must have the same size");
-    std::memcpy(&f, &u, sizeof(u));
-    return f;
-#endif
-  }
 };
 
 template <>