H3cJP/source-engine
1
0
Fork 0
mirror of https://github.com/nillerusr/source-engine.git synced 2026-08-11 11:19:09 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

First steps

Browse Source
This commit is contained in:
JusicP
2020-11-19 22:37:27 +02:00
parent 6920a3bd08
commit 3434111923
1341 changed files with 883624 additions and 46746 deletions
Download Patch File Download Diff File Expand all files Collapse all files
public/mathlib/mathlib.h
+6 -2
View File
@@ -454,7 +454,7 @@ void inline SinCos( float radians, float *sine, float *cosine )
fstp DWORD PTR [edx]
fstp DWORD PTR [eax]
}
#elif defined( PLATFORM_WINDOWS_PC64 )
#elif defined( PLATFORM_WINDOWS_PC64 ) || defined(__arm__)
*sine = sin( radians );
*cosine = cos( radians );
#elif defined( POSIX )
@@ -1217,6 +1217,8 @@ FORCEINLINE int RoundFloatToInt(float f)
};
flResult = __fctiw( f );
return pResult[1];
#elif defined(__arm__)
return (int)(f + 0.5f);
#else
#error Unknown architecture
#endif
@@ -1246,8 +1248,10 @@ FORCEINLINE unsigned long RoundFloatToUnsignedLong(float f)
flResult = __fctiw( f );
Assert( pIntResult[1] >= 0 );
return pResult[1];
#elif defined(__arm__)
return (unsigned long)(f + 0.5f);
#else // !X360
#if defined( PLATFORM_WINDOWS_PC64 )
uint nRet = ( uint ) f;
if ( nRet & 1 )
public/mathlib/ssemath.h
+16 -8
View File
@@ -8,6 +8,8 @@
#if defined( _X360 )
#include <xboxmath.h>
#elif defined(__arm__)
#include <SSE2NEON.h>
#else
#include <xmmintrin.h>
#endif
@@ -15,7 +17,7 @@
#include <mathlib/vector.h>
#include <mathlib/mathlib.h>
#if defined(GNUC)
#if defined(GNUC) && defined(__arm__) // HACK: implement NEON later
#define USE_STDC_FOR_SIMD 0
#else
#define USE_STDC_FOR_SIMD 0
@@ -879,10 +881,13 @@ FORCEINLINE fltx4 FindHighestSIMD3( const fltx4 & a )
// like this.
FORCEINLINE void ConvertStoreAsIntsSIMD(intx4 * RESTRICT pDest, const fltx4 &vSrc)
{
(*pDest)[0] = SubFloat(vSrc, 0);
(*pDest)[1] = SubFloat(vSrc, 1);
(*pDest)[2] = SubFloat(vSrc, 2);
(*pDest)[3] = SubFloat(vSrc, 3);
__m64 bottom = _mm_cvttps_pi32( vSrc );
__m64 top = _mm_cvttps_pi32( _mm_movehl_ps(vSrc,vSrc) );
*reinterpret_cast<__m64 *>(&(*pDest)[0]) = bottom;
*reinterpret_cast<__m64 *>(&(*pDest)[2]) = top;
_mm_empty();
}
// ------------------------------------
@@ -2407,13 +2412,16 @@ FORCEINLINE i32x4 IntShiftLeftWordSIMD(const i32x4 &vSrcA, const i32x4 &vSrcB)
// like this.
FORCEINLINE void ConvertStoreAsIntsSIMD(intx4 * RESTRICT pDest, const fltx4 &vSrc)
{
#if defined( COMPILER_MSVC64 )
#ifdef __arm__
(*pDest)[0] = (int)vSrc[0];
(*pDest)[1] = (int)vSrc[1];
(*pDest)[2] = (int)vSrc[2];
(*pDest)[3] = (int)vSrc[3];
#elif defined( COMPILER_MSVC64 )
(*pDest)[0] = SubFloat( vSrc, 0 );
(*pDest)[1] = SubFloat( vSrc, 1 );
(*pDest)[2] = SubFloat( vSrc, 2 );
(*pDest)[3] = SubFloat( vSrc, 3 );
#else
__m64 bottom = _mm_cvttps_pi32( vSrc );
__m64 top = _mm_cvttps_pi32( _mm_movehl_ps(vSrc,vSrc) );
public/mathlib/vector.h
+4 -4
View File
@@ -22,7 +22,7 @@
// For rand(). We really need a library!
#include <stdlib.h>
#ifndef _X360
#if !defined(_X360) && !defined(__arm__)
// For MMX intrinsics
#include <xmmintrin.h>
#endif
@@ -210,7 +210,7 @@ private:
FORCEINLINE void NetworkVarConstruct( Vector &v ) { v.Zero(); }
#define USE_M64S ( ( !defined( _X360 ) ) )
#define USE_M64S ( ( !defined( _X360 ) && !defined(__arm__) ) )
@@ -227,7 +227,7 @@ public:
void Init(short ix = 0, short iy = 0, short iz = 0, short iw = 0 );
#ifdef USE_M64S
#if USE_M64S
__m64 &AsM64() { return *(__m64*)&x; }
const __m64 &AsM64() const { return *(const __m64*)&x; }
#endif
@@ -284,7 +284,7 @@ public:
// Initialization
void Init(int ix = 0, int iy = 0, int iz = 0, int iw = 0 );
#ifdef USE_M64S
#if USE_M64S
__m64 &AsM64() { return *(__m64*)&x; }
const __m64 &AsM64() const { return *(const __m64*)&x; }
#endif
public/mathlib/vector4d.h
+10 -4
View File
@@ -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( _X360 ) && !defined(__arm__)
#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
@@ -614,7 +616,9 @@ 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 = 0.0f;
#elif !defined( _X360 )
this->AsM128() = _mm_set1_ps( 0.0f );
#else
this->AsM128() = __vspltisw( 0 );
@@ -663,8 +667,10 @@ inline void Vector4DWeightMAD( vec_t w, Vector4DAligned const& vInA, Vector4DAli
inline void Vector4DWeightMADSSE( vec_t w, Vector4DAligned const& vInA, Vector4DAligned& vOutA, Vector4DAligned const& vInB, Vector4DAligned& vOutB )
{
Assert( vInA.IsValid() && vInB.IsValid() && IsFinite(w) );
#if !defined( _X360 )
#if defined(__arm__)
Vector4DWeightMAD( w, vInA, vOutA, vInB, vOutB );
#elif !defined( _X360 )
// Replicate scalar float out to 4 components
__m128 packed = _mm_set1_ps( w );