build: arm target support

public/mathlib/mathlib.h

@@ -458,11 +458,7 @@ void inline SinCos( float radians, float *sine, float *cosine )
 	*sine = sin( radians );
 	*cosine = cos( radians );
 #elif defined( POSIX )
-	double __cosr, __sinr;
-	__asm ("fsincos" : "=t" (__cosr), "=u" (__sinr) : "0" (radians));
-
-  	*sine = __sinr;
-  	*cosine = __cosr;
+	sincosf(radians, sine, cosine);
 #endif
 }
 
@@ -1217,6 +1213,8 @@ FORCEINLINE int RoundFloatToInt(float f)
 	};
 	flResult = __fctiw( f );
 	return pResult[1];
+#elif defined (__arm__)
+        return (int)(f + 0.5f);
 #else
 #error Unknown architecture
 #endif
@@ -1247,8 +1245,9 @@ FORCEINLINE unsigned long RoundFloatToUnsignedLong(float f)
 	Assert( pIntResult[1] >= 0 );
 	return pResult[1];
 #else  // !X360
-	
-#if defined( PLATFORM_WINDOWS_PC64 )
+#ifdef __arm__
+        return (unsigned long)(f + 0.5f);
+#elif defined( PLATFORM_WINDOWS_PC64 )
 	uint nRet = ( uint ) f;
 	if ( nRet & 1 )
 	{

public/mathlib/ssemath.h

@@ -8,6 +8,8 @@
 
 #if defined( _X360 )
 #include <xboxmath.h>
+#elif defined(__arm__)
+#include "sse2neon.h"
 #else
 #include <xmmintrin.h>
 #endif
@@ -21,7 +23,7 @@
 #define USE_STDC_FOR_SIMD 0
 #endif
 
-#if (!defined(_X360) && (USE_STDC_FOR_SIMD == 0))
+#if (!defined (__arm__) && !defined(_X360) && (USE_STDC_FOR_SIMD == 0))
 #define _SSE1 1
 #endif
 

public/mathlib/vector.h

@@ -22,7 +22,8 @@
 // For rand(). We really need a library!
 #include <stdlib.h>
 
-#ifndef _X360
+#if defined(__SSE__) || defined(_M_IX86_FP)
+#define USE_SSE
 // For MMX intrinsics
 #include <xmmintrin.h>
 #endif
@@ -209,10 +210,9 @@ private:
 
 FORCEINLINE void NetworkVarConstruct( Vector &v ) { v.Zero(); }
 
-
-#define USE_M64S ( ( !defined( _X360 ) ) )
-
-
+#ifdef USE_SSE
+#define USE_M64S
+#endif
 
 //=========================================================
 // 4D Short Vector (aligned on 8-byte boundary)

public/mathlib/vector4d.h

@@ -16,7 +16,7 @@
 #include <math.h>
 #include <stdlib.h>		// For rand(). We really need a library!
 #include <float.h>
-#if !defined( _X360 )
+#if defined(__SSE__) || defined(_M_IX86_FP)
 #include <xmmintrin.h>	// For SSE
 #endif
 #include "basetypes.h"	// For vec_t, put this somewhere else?
@@ -141,8 +141,10 @@ public:
 	inline void Set( vec_t X, vec_t Y, vec_t Z, vec_t W );
 	inline void InitZero( void );
 
+#ifndef __arm__
 	inline __m128 &AsM128() { return *(__m128*)&x; }
 	inline const __m128 &AsM128() const { return *(const __m128*)&x; } 
+#endif
 
 private:
 	// No copy constructors allowed if we're in optimal mode
@@ -613,8 +615,10 @@ inline void Vector4DAligned::Set( vec_t X, vec_t Y, vec_t Z, vec_t W )
 }
 
 inline void Vector4DAligned::InitZero( void )
-{ 
-#if !defined( _X360 )
+{
+#if defined (__arm__)
+	x = y = z = w = 0;
+#elif !defined( _X360 )
 	this->AsM128() = _mm_set1_ps( 0.0f );
 #else
 	this->AsM128() = __vspltisw( 0 );
@@ -625,7 +629,7 @@ inline void Vector4DAligned::InitZero( void )
 inline void Vector4DMultiplyAligned( Vector4DAligned const& a, Vector4DAligned const& b, Vector4DAligned& c )
 {
 	Assert( a.IsValid() && b.IsValid() );
-#if !defined( _X360 )
+#if !defined( _X360 ) || defined (__arm__)
 	c.x = a.x * b.x;
 	c.y = a.y * b.y;
 	c.z = a.z * b.z;
@@ -639,7 +643,7 @@ inline void Vector4DWeightMAD( vec_t w, Vector4DAligned const& vInA, Vector4DAli
 {
 	Assert( vInA.IsValid() && vInB.IsValid() && IsFinite(w) );
 
-#if !defined( _X360 )
+#if !defined( _X360 ) || defined (__arm__)
 	vOutA.x += vInA.x * w;
 	vOutA.y += vInA.y * w;
 	vOutA.z += vInA.z * w;
@@ -660,6 +664,7 @@ inline void Vector4DWeightMAD( vec_t w, Vector4DAligned const& vInA, Vector4DAli
 #endif
 }
 
+#ifndef __arm__
 inline void Vector4DWeightMADSSE( vec_t w, Vector4DAligned const& vInA, Vector4DAligned& vOutA, Vector4DAligned const& vInB, Vector4DAligned& vOutB )
 {
 	Assert( vInA.IsValid() && vInB.IsValid() && IsFinite(w) );
@@ -681,6 +686,7 @@ inline void Vector4DWeightMADSSE( vec_t w, Vector4DAligned const& vInA, Vector4D
 	vOutB.AsM128() = __vmaddfp( vInB.AsM128(), temp, vOutB.AsM128() );
 #endif
 }
+#endif
 
 #endif // VECTOR4D_H