feat: Add float overloads to WWMath to ensure deterministic single-precision math results

Author: Okladnoj

Core/Libraries/Source/WWVegas/WWMath/wwmath.h

@@ -40,6 +40,7 @@
 #include <math.h>
 #include <float.h>
 #include <assert.h>
+#include "Lib/BaseDefines.h"
 
 #if USE_DETERMINISTIC_MATH
 #include "gmath.h"
@@ -139,6 +140,21 @@ static WWINLINE float  Sinf_Legacy(float val);
 static WWINLINE double Tan(double x);
 static WWINLINE float  Tanf(float x);
 
+	// ── FLOAT OVERLOADS FOR DETERMINISTIC MATH ──
+	// Prevent automatic compiler promotion of float arguments to double-precision variants.
+	// Single-precision math in GameMath (gm_*f) is guaranteed to be cross-platform bit-identical,
+	// whereas double-precision math (gm_*) can diverge by 1 ULP due to FPU precision differences (x87 vs NEON).
+	// Example fact: at Frame 1 on Akas Magic, Object 312 ROTATE dy=446C04C6 dx=C3A4DDB0 had:
+	// WWMath::Atan2(float, float) -> resolved to double Atan2 -> gm_atan2(double, double)
+	// resulting in Mac: 3FF4128C vs Win: 3FF4128B (1 ULP mismatch).
+	static WWINLINE float  Sin(float val);
+	static WWINLINE float  Cos(float val);
+	static WWINLINE float  Tan(float x);
+	static WWINLINE float  Atan(float x);
+	static WWINLINE float  Atan2(float x, float y);
+	static WWINLINE float  Asin(float x);
+	static WWINLINE float  Acos(float x);
+
 static WWINLINE double Cosh(double x);
 static WWINLINE float  Coshf(float x);
 static WWINLINE double Sinh(double x);
@@ -484,6 +500,72 @@ WWINLINE float WWMath::Atan2f(float x, float y)
 #endif
 }
 
+// ── FLOAT OVERLOADS FOR DETERMINISTIC MATH ──
+// Inline implementations routing float calls directly to Single-precision GameMath functions.
+
+WWINLINE float WWMath::Sin(float val)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_sinf(val);
+#else
+	return sinf(val);
+#endif
+}
+
+WWINLINE float WWMath::Cos(float val)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_cosf(val);
+#else
+	return cosf(val);
+#endif
+}
+
+WWINLINE float WWMath::Tan(float x)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_tanf(x);
+#else
+	return tanf(x);
+#endif
+}
+
+WWINLINE float WWMath::Atan(float x)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_atanf(x);
+#else
+	return atanf(x);
+#endif
+}
+
+WWINLINE float WWMath::Atan2(float x, float y)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_atan2f(x, y);
+#else
+	return atan2f(x, y);
+#endif
+}
+
+WWINLINE float WWMath::Asin(float x)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_asinf(x);
+#else
+	return asinf(x);
+#endif
+}
+
+WWINLINE float WWMath::Acos(float x)
+{
+#if USE_DETERMINISTIC_MATH
+	return gm_acosf(x);
+#else
+	return acosf(x);
+#endif
+}
+
 WWINLINE float WWMath::Fast_Cos(float val)
 {
 	val+=(WWMATH_PI * 0.5f);