dhewm/dhewm3-sdk
0
0
Fork 0
mirror of https://github.com/dhewm/dhewm3-sdk.git synced 2024-11-25 14:01:10 +00:00
Code Issues Projects Releases Packages Wiki Activity

Fix ./idlib/math/ for x86_64

Browse source 
sizeof(float) != sizeof(long) on this ABI.
This commit is contained in:
dhewg 2011-12-01 16:31:11 +01:00 committed by Daniel Gibson
parent bca35dec66
commit dc86a8a02c
5 changed files with 73 additions and 73 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

38
idlib/math/Math.h
View file

@ -63,17 +63,17 @@ If you have questions concerning this license or the applicable additional terms
#define ANGLE2BYTE(x) ( idMath::FtoiFast( (x) * 256.0f / 360.0f ) & 255 ) #define ANGLE2BYTE(x) ( idMath::FtoiFast( (x) * 256.0f / 360.0f ) & 255 )
#define BYTE2ANGLE(x) ( (x) * ( 360.0f / 256.0f ) ) #define BYTE2ANGLE(x) ( (x) * ( 360.0f / 256.0f ) )
#define FLOATSIGNBITSET(f) ((*(const unsigned long *)&(f)) >> 31) #define FLOATSIGNBITSET(f) ((*(const unsigned int *)&(f)) >> 31)
#define FLOATSIGNBITNOTSET(f) ((~(*(const unsigned long *)&(f))) >> 31) #define FLOATSIGNBITNOTSET(f) ((~(*(const unsigned int *)&(f))) >> 31)
#define FLOATNOTZERO(f) ((*(const unsigned long *)&(f)) & ~(1<<31) ) #define FLOATNOTZERO(f) ((*(const unsigned int *)&(f)) & ~(1<<31) )
#define INTSIGNBITSET(i) (((const unsigned long)(i)) >> 31) #define INTSIGNBITSET(i) (((const unsigned int)(i)) >> 31)
#define INTSIGNBITNOTSET(i) ((~((const unsigned long)(i))) >> 31) #define INTSIGNBITNOTSET(i) ((~((const unsigned int)(i))) >> 31)
#define FLOAT_IS_NAN(x) (((*(const unsigned long *)&x) & 0x7f800000) == 0x7f800000) #define FLOAT_IS_NAN(x) (((*(const unsigned int *)&x) & 0x7f800000) == 0x7f800000)
#define FLOAT_IS_INF(x) (((*(const unsigned long *)&x) & 0x7fffffff) == 0x7f800000) #define FLOAT_IS_INF(x) (((*(const unsigned int *)&x) & 0x7fffffff) == 0x7f800000)
#define FLOAT_IS_IND(x) ((*(const unsigned long *)&x) == 0xffc00000) #define FLOAT_IS_IND(x) ((*(const unsigned int *)&x) == 0xffc00000)
#define FLOAT_IS_DENORMAL(x) (((*(const unsigned long *)&x) & 0x7f800000) == 0x00000000 && \ #define FLOAT_IS_DENORMAL(x) (((*(const unsigned int *)&x) & 0x7f800000) == 0x00000000 && \
((*(const unsigned long *)&x) & 0x007fffff) != 0x00000000 ) ((*(const unsigned int *)&x) & 0x007fffff) != 0x00000000 )
#define IEEE_FLT_MANTISSA_BITS 23 #define IEEE_FLT_MANTISSA_BITS 23
#define IEEE_FLT_EXPONENT_BITS 8 #define IEEE_FLT_EXPONENT_BITS 8
@ -183,8 +183,8 @@ public:
static float Rint( float f ); // returns the nearest integer static float Rint( float f ); // returns the nearest integer
static int Ftoi( float f ); // float to int conversion static int Ftoi( float f ); // float to int conversion
static int FtoiFast( float f ); // fast float to int conversion but uses current FPU round mode (default round nearest) static int FtoiFast( float f ); // fast float to int conversion but uses current FPU round mode (default round nearest)
static unsigned long Ftol( float f ); // float to long conversion static unsigned int Ftol( float f ); // float to int conversion
static unsigned long FtolFast( float ); // fast float to long conversion but uses current FPU round mode (default round nearest) static unsigned int FtolFast( float ); // fast float to int conversion but uses current FPU round mode (default round nearest)
static signed char ClampChar( int i ); static signed char ClampChar( int i );
static signed short ClampShort( int i ); static signed short ClampShort( int i );
@ -238,11 +238,11 @@ private:
ID_INLINE float idMath::RSqrt( float x ) { ID_INLINE float idMath::RSqrt( float x ) {
long i; int i;
float y, r; float y, r;
y = x * 0.5f; y = x * 0.5f;
i = *reinterpret_cast<long *>( &x ); i = *reinterpret_cast<int *>( &x );
i = 0x5f3759df - ( i >> 1 ); i = 0x5f3759df - ( i >> 1 );
r = *reinterpret_cast<float *>( &i ); r = *reinterpret_cast<float *>( &i );
r = r * ( 1.5f - r * r * y ); r = r * ( 1.5f - r * r * y );
@ -824,14 +824,14 @@ ID_INLINE int idMath::FtoiFast( float f ) {
#endif #endif
} }
ID_INLINE unsigned long idMath::Ftol( float f ) { ID_INLINE unsigned int idMath::Ftol( float f ) {
return (unsigned long) f; return (unsigned int) f;
} }
ID_INLINE unsigned long idMath::FtolFast( float f ) { ID_INLINE unsigned int idMath::FtolFast( float f ) {
#ifdef _WIN32 #ifdef _WIN32
// FIXME: this overflows on 31bits still .. same as FtoiFast // FIXME: this overflows on 31bits still .. same as FtoiFast
unsigned long i; unsigned int i;
__asm fld f __asm fld f
__asm fistp i // use default rouding mode (round nearest) __asm fistp i // use default rouding mode (round nearest)
return i; return i;
@ -854,7 +854,7 @@ ID_INLINE unsigned long idMath::FtolFast( float f ) {
: "m" (f) ); : "m" (f) );
return i; return i;
#else #else
return (unsigned long) f; return (unsigned int) f;
#endif #endif
} }

4
idlib/math/Simd_AltiVec.cpp
View file

@ -68,8 +68,8 @@ If you have questions concerning this license or the applicable additional terms
#define IDVEC4_OFFSET 4 #define IDVEC4_OFFSET 4
// Alignment tests // Alignment tests
#define IS_16BYTE_ALIGNED( x ) ( ( (unsigned long)&x & 0x0F ) == 0 ) #define IS_16BYTE_ALIGNED( x ) ( ( (unsigned int)&x & 0x0F ) == 0 )
#define NOT_16BYTE_ALIGNED( x ) ( ( (unsigned long)&x & 0x0F) != 0 ) #define NOT_16BYTE_ALIGNED( x ) ( ( (unsigned int)&x & 0x0F) != 0 )
// Aligned storing floats // Aligned storing floats
#define ALIGNED_STORE2( ADDR, V0, V1 ) \ #define ALIGNED_STORE2( ADDR, V0, V1 ) \

8
idlib/math/Simd_Generic.cpp
View file

@ -2499,7 +2499,7 @@ void VPCALL idSIMD_Generic::DeriveTangents( idPlane *planes, idDrawVert *verts,
idPlane *planesPtr = planes; idPlane *planesPtr = planes;
for ( i = 0; i < numIndexes; i += 3 ) { for ( i = 0; i < numIndexes; i += 3 ) {
idDrawVert *a, *b, *c; idDrawVert *a, *b, *c;
unsigned long signBit; unsigned int signBit;
float d0[5], d1[5], f, area; float d0[5], d1[5], f, area;
idVec3 n, t0, t1; idVec3 n, t0, t1;
@ -2540,7 +2540,7 @@ void VPCALL idSIMD_Generic::DeriveTangents( idPlane *planes, idDrawVert *verts,
// area sign bit // area sign bit
area = d0[3] * d1[4] - d0[4] * d1[3]; area = d0[3] * d1[4] - d0[4] * d1[3];
signBit = ( *(unsigned long *)&area ) & ( 1 << 31 ); signBit = ( *(unsigned int *)&area ) & ( 1 << 31 );
// first tangent // first tangent
t0[0] = d0[0] * d1[4] - d0[4] * d1[0]; t0[0] = d0[0] * d1[4] - d0[4] * d1[0];
@ -2548,7 +2548,7 @@ void VPCALL idSIMD_Generic::DeriveTangents( idPlane *planes, idDrawVert *verts,
t0[2] = d0[2] * d1[4] - d0[4] * d1[2]; t0[2] = d0[2] * d1[4] - d0[4] * d1[2];
f = idMath::RSqrt( t0.x * t0.x + t0.y * t0.y + t0.z * t0.z ); f = idMath::RSqrt( t0.x * t0.x + t0.y * t0.y + t0.z * t0.z );
*(unsigned long *)&f ^= signBit; *(unsigned int *)&f ^= signBit;
t0.x *= f; t0.x *= f;
t0.y *= f; t0.y *= f;
@ -2560,7 +2560,7 @@ void VPCALL idSIMD_Generic::DeriveTangents( idPlane *planes, idDrawVert *verts,
t1[2] = d0[3] * d1[2] - d0[2] * d1[3]; t1[2] = d0[3] * d1[2] - d0[2] * d1[3];
f = idMath::RSqrt( t1.x * t1.x + t1.y * t1.y + t1.z * t1.z ); f = idMath::RSqrt( t1.x * t1.x + t1.y * t1.y + t1.z * t1.z );
*(unsigned long *)&f ^= signBit; *(unsigned int *)&f ^= signBit;
t1.x *= f; t1.x *= f;
t1.y *= f; t1.y *= f;

88
idlib/math/Simd_SSE.cpp
View file

@ -1017,16 +1017,16 @@ void VPCALL idSIMD_SSE::Dot( float *dst, const idVec3 &constant, const idPlane *
ALIGN8_INIT1( unsigned short SIMD_W_zero, 0 ); ALIGN8_INIT1( unsigned short SIMD_W_zero, 0 );
ALIGN8_INIT1( unsigned short SIMD_W_maxShort, 1<<15 ); ALIGN8_INIT1( unsigned short SIMD_W_maxShort, 1<<15 );
ALIGN4_INIT1( unsigned long SIMD_DW_mat2quatShuffle0, (3<<0)|(2<<8)|(1<<16)|(0<<24) ); ALIGN4_INIT1( unsigned int SIMD_DW_mat2quatShuffle0, (3<<0)|(2<<8)|(1<<16)|(0<<24) );
ALIGN4_INIT1( unsigned long SIMD_DW_mat2quatShuffle1, (0<<0)|(1<<8)|(2<<16)|(3<<24) ); ALIGN4_INIT1( unsigned int SIMD_DW_mat2quatShuffle1, (0<<0)|(1<<8)|(2<<16)|(3<<24) );
ALIGN4_INIT1( unsigned long SIMD_DW_mat2quatShuffle2, (1<<0)|(0<<8)|(3<<16)|(2<<24) ); ALIGN4_INIT1( unsigned int SIMD_DW_mat2quatShuffle2, (1<<0)|(0<<8)|(3<<16)|(2<<24) );
ALIGN4_INIT1( unsigned long SIMD_DW_mat2quatShuffle3, (2<<0)|(3<<8)|(0<<16)|(1<<24) ); ALIGN4_INIT1( unsigned int SIMD_DW_mat2quatShuffle3, (2<<0)|(3<<8)|(0<<16)|(1<<24) );
ALIGN4_INIT4( unsigned long SIMD_SP_singleSignBitMask, (unsigned long) ( 1 << 31 ), 0, 0, 0 ); ALIGN4_INIT4( unsigned int SIMD_SP_singleSignBitMask, (unsigned int) ( 1 << 31 ), 0, 0, 0 );
ALIGN4_INIT1( unsigned long SIMD_SP_signBitMask, (unsigned long) ( 1 << 31 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_signBitMask, (unsigned int) ( 1 << 31 ) );
ALIGN4_INIT1( unsigned long SIMD_SP_absMask, (unsigned long) ~( 1 << 31 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_absMask, (unsigned int) ~( 1 << 31 ) );
ALIGN4_INIT1( unsigned long SIMD_SP_infinityMask, (unsigned long) ~( 1 << 23 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_infinityMask, (unsigned int) ~( 1 << 23 ) );
ALIGN4_INIT1( unsigned long SIMD_SP_not, 0xFFFFFFFF ); ALIGN4_INIT1( unsigned int SIMD_SP_not, 0xFFFFFFFF );
ALIGN4_INIT1( float SIMD_SP_zero, 0.0f ); ALIGN4_INIT1( float SIMD_SP_zero, 0.0f );
ALIGN4_INIT1( float SIMD_SP_half, 0.5f ); ALIGN4_INIT1( float SIMD_SP_half, 0.5f );
@ -1950,7 +1950,7 @@ float SSE_ATan( float y, float x ) {
if ( fabs( y ) > fabs( x ) ) { if ( fabs( y ) > fabs( x ) ) {
a = -x / y; a = -x / y;
d = idMath::HALF_PI; d = idMath::HALF_PI;
*((unsigned long *)&d) ^= ( *((unsigned long *)&x) ^ *((unsigned long *)&y) ) & (1<<31); *((unsigned int *)&d) ^= ( *((unsigned int *)&x) ^ *((unsigned int *)&y) ) & (1<<31);
} else { } else {
a = y / x; a = y / x;
d = 0.0f; d = 0.0f;
@ -11527,7 +11527,7 @@ void VPCALL idSIMD_SSE::BlendJoints( idJointQuat *joints, const idJointQuat *ble
ALIGN16( float omega1[4] ); ALIGN16( float omega1[4] );
ALIGN16( float scale0[4] ); ALIGN16( float scale0[4] );
ALIGN16( float scale1[4] ); ALIGN16( float scale1[4] );
ALIGN16( unsigned long signBit[4] ); ALIGN16( unsigned int signBit[4] );
cosom[0] = jointQuat0[0] * blendQuat0[0]; cosom[0] = jointQuat0[0] * blendQuat0[0];
cosom[1] = jointQuat0[1] * blendQuat0[1]; cosom[1] = jointQuat0[1] * blendQuat0[1];
@ -11549,15 +11549,15 @@ void VPCALL idSIMD_SSE::BlendJoints( idJointQuat *joints, const idJointQuat *ble
cosom[2] += jointQuat3[2] * blendQuat3[2]; cosom[2] += jointQuat3[2] * blendQuat3[2];
cosom[3] += jointQuat3[3] * blendQuat3[3]; cosom[3] += jointQuat3[3] * blendQuat3[3];
signBit[0] = (*(unsigned long *)&cosom[0]) & ( 1 << 31 ); signBit[0] = (*(unsigned int *)&cosom[0]) & ( 1 << 31 );
signBit[1] = (*(unsigned long *)&cosom[1]) & ( 1 << 31 ); signBit[1] = (*(unsigned int *)&cosom[1]) & ( 1 << 31 );
signBit[2] = (*(unsigned long *)&cosom[2]) & ( 1 << 31 ); signBit[2] = (*(unsigned int *)&cosom[2]) & ( 1 << 31 );
signBit[3] = (*(unsigned long *)&cosom[3]) & ( 1 << 31 ); signBit[3] = (*(unsigned int *)&cosom[3]) & ( 1 << 31 );
(*(unsigned long *)&cosom[0]) ^= signBit[0]; (*(unsigned int *)&cosom[0]) ^= signBit[0];
(*(unsigned long *)&cosom[1]) ^= signBit[1]; (*(unsigned int *)&cosom[1]) ^= signBit[1];
(*(unsigned long *)&cosom[2]) ^= signBit[2]; (*(unsigned int *)&cosom[2]) ^= signBit[2];
(*(unsigned long *)&cosom[3]) ^= signBit[3]; (*(unsigned int *)&cosom[3]) ^= signBit[3];
scale0[0] = 1.0f - cosom[0] * cosom[0]; scale0[0] = 1.0f - cosom[0] * cosom[0];
scale0[1] = 1.0f - cosom[1] * cosom[1]; scale0[1] = 1.0f - cosom[1] * cosom[1];
@ -11604,10 +11604,10 @@ void VPCALL idSIMD_SSE::BlendJoints( idJointQuat *joints, const idJointQuat *ble
scale1[2] = SSE_SinZeroHalfPI( omega1[2] ) * sinom[2]; scale1[2] = SSE_SinZeroHalfPI( omega1[2] ) * sinom[2];
scale1[3] = SSE_SinZeroHalfPI( omega1[3] ) * sinom[3]; scale1[3] = SSE_SinZeroHalfPI( omega1[3] ) * sinom[3];
(*(unsigned long *)&scale1[0]) ^= signBit[0]; (*(unsigned int *)&scale1[0]) ^= signBit[0];
(*(unsigned long *)&scale1[1]) ^= signBit[1]; (*(unsigned int *)&scale1[1]) ^= signBit[1];
(*(unsigned long *)&scale1[2]) ^= signBit[2]; (*(unsigned int *)&scale1[2]) ^= signBit[2];
(*(unsigned long *)&scale1[3]) ^= signBit[3]; (*(unsigned int *)&scale1[3]) ^= signBit[3];
jointQuat0[0] = scale0[0] * jointQuat0[0] + scale1[0] * blendQuat0[0]; jointQuat0[0] = scale0[0] * jointQuat0[0] + scale1[0] * blendQuat0[0];
jointQuat0[1] = scale0[1] * jointQuat0[1] + scale1[1] * blendQuat0[1]; jointQuat0[1] = scale0[1] * jointQuat0[1] + scale1[1] * blendQuat0[1];
@ -11663,13 +11663,13 @@ void VPCALL idSIMD_SSE::BlendJoints( idJointQuat *joints, const idJointQuat *ble
float omega; float omega;
float scale0; float scale0;
float scale1; float scale1;
unsigned long signBit; unsigned int signBit;
cosom = jointQuat.x * blendQuat.x + jointQuat.y * blendQuat.y + jointQuat.z * blendQuat.z + jointQuat.w * blendQuat.w; cosom = jointQuat.x * blendQuat.x + jointQuat.y * blendQuat.y + jointQuat.z * blendQuat.z + jointQuat.w * blendQuat.w;
signBit = (*(unsigned long *)&cosom) & ( 1 << 31 ); signBit = (*(unsigned int *)&cosom) & ( 1 << 31 );
(*(unsigned long *)&cosom) ^= signBit; (*(unsigned int *)&cosom) ^= signBit;
scale0 = 1.0f - cosom * cosom; scale0 = 1.0f - cosom * cosom;
scale0 = ( scale0 <= 0.0f ) ? SIMD_SP_tiny[0] : scale0; scale0 = ( scale0 <= 0.0f ) ? SIMD_SP_tiny[0] : scale0;
@ -11678,7 +11678,7 @@ void VPCALL idSIMD_SSE::BlendJoints( idJointQuat *joints, const idJointQuat *ble
scale0 = idMath::Sin16( ( 1.0f - lerp ) * omega ) * sinom; scale0 = idMath::Sin16( ( 1.0f - lerp ) * omega ) * sinom;
scale1 = idMath::Sin16( lerp * omega ) * sinom; scale1 = idMath::Sin16( lerp * omega ) * sinom;
(*(unsigned long *)&scale1) ^= signBit; (*(unsigned int *)&scale1) ^= signBit;
jointQuat.x = scale0 * jointQuat.x + scale1 * blendQuat.x; jointQuat.x = scale0 * jointQuat.x + scale1 * blendQuat.x;
jointQuat.y = scale0 * jointQuat.y + scale1 * blendQuat.y; jointQuat.y = scale0 * jointQuat.y + scale1 * blendQuat.y;
@ -13664,7 +13664,7 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
for ( i = 0; i <= numIndexes - 12; i += 12 ) { for ( i = 0; i <= numIndexes - 12; i += 12 ) {
idDrawVert *a, *b, *c; idDrawVert *a, *b, *c;
ALIGN16( unsigned long signBit[4] ); ALIGN16( unsigned int signBit[4] );
ALIGN16( float d0[4] ); ALIGN16( float d0[4] );
ALIGN16( float d1[4] ); ALIGN16( float d1[4] );
ALIGN16( float d2[4] ); ALIGN16( float d2[4] );
@ -13967,10 +13967,10 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
tmp[2] -= d4[2] * d8[2]; tmp[2] -= d4[2] * d8[2];
tmp[3] -= d4[3] * d8[3]; tmp[3] -= d4[3] * d8[3];
signBit[0] = ( *(unsigned long *)&tmp[0] ) & ( 1 << 31 ); signBit[0] = ( *(unsigned int *)&tmp[0] ) & ( 1 << 31 );
signBit[1] = ( *(unsigned long *)&tmp[1] ) & ( 1 << 31 ); signBit[1] = ( *(unsigned int *)&tmp[1] ) & ( 1 << 31 );
signBit[2] = ( *(unsigned long *)&tmp[2] ) & ( 1 << 31 ); signBit[2] = ( *(unsigned int *)&tmp[2] ) & ( 1 << 31 );
signBit[3] = ( *(unsigned long *)&tmp[3] ) & ( 1 << 31 ); signBit[3] = ( *(unsigned int *)&tmp[3] ) & ( 1 << 31 );
// first tangent // first tangent
t0[0] = d0[0] * d9[0]; t0[0] = d0[0] * d9[0];
@ -14023,10 +14023,10 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
tmp[2] = idMath::RSqrt( tmp[2] ); tmp[2] = idMath::RSqrt( tmp[2] );
tmp[3] = idMath::RSqrt( tmp[3] ); tmp[3] = idMath::RSqrt( tmp[3] );
*(unsigned long *)&tmp[0] ^= signBit[0]; *(unsigned int *)&tmp[0] ^= signBit[0];
*(unsigned long *)&tmp[1] ^= signBit[1]; *(unsigned int *)&tmp[1] ^= signBit[1];
*(unsigned long *)&tmp[2] ^= signBit[2]; *(unsigned int *)&tmp[2] ^= signBit[2];
*(unsigned long *)&tmp[3] ^= signBit[3]; *(unsigned int *)&tmp[3] ^= signBit[3];
t0[0] *= tmp[0]; t0[0] *= tmp[0];
t0[1] *= tmp[1]; t0[1] *= tmp[1];
@ -14094,10 +14094,10 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
tmp[2] = idMath::RSqrt( tmp[2] ); tmp[2] = idMath::RSqrt( tmp[2] );
tmp[3] = idMath::RSqrt( tmp[3] ); tmp[3] = idMath::RSqrt( tmp[3] );
*(unsigned long *)&tmp[0] ^= signBit[0]; *(unsigned int *)&tmp[0] ^= signBit[0];
*(unsigned long *)&tmp[1] ^= signBit[1]; *(unsigned int *)&tmp[1] ^= signBit[1];
*(unsigned long *)&tmp[2] ^= signBit[2]; *(unsigned int *)&tmp[2] ^= signBit[2];
*(unsigned long *)&tmp[3] ^= signBit[3]; *(unsigned int *)&tmp[3] ^= signBit[3];
t3[0] *= tmp[0]; t3[0] *= tmp[0];
t3[1] *= tmp[1]; t3[1] *= tmp[1];
@ -14220,7 +14220,7 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
for ( ; i < numIndexes; i += 3 ) { for ( ; i < numIndexes; i += 3 ) {
idDrawVert *a, *b, *c; idDrawVert *a, *b, *c;
ALIGN16( unsigned long signBit[4] ); ALIGN16( unsigned int signBit[4] );
float d0, d1, d2, d3, d4; float d0, d1, d2, d3, d4;
float d5, d6, d7, d8, d9; float d5, d6, d7, d8, d9;
float n0, n1, n2; float n0, n1, n2;
@ -14446,7 +14446,7 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
// area sign bit // area sign bit
tmp = d3 * d9 - d4 * d8; tmp = d3 * d9 - d4 * d8;
signBit[0] = ( *(unsigned long *)&tmp ) & ( 1 << 31 ); signBit[0] = ( *(unsigned int *)&tmp ) & ( 1 << 31 );
// first tangent // first tangent
t0 = d0 * d9 - d4 * d5; t0 = d0 * d9 - d4 * d5;
@ -14454,7 +14454,7 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
t2 = d2 * d9 - d4 * d7; t2 = d2 * d9 - d4 * d7;
tmp = idMath::RSqrt( t0 * t0 + t1 * t1 + t2 * t2 ); tmp = idMath::RSqrt( t0 * t0 + t1 * t1 + t2 * t2 );
*(unsigned long *)&tmp ^= signBit[0]; *(unsigned int *)&tmp ^= signBit[0];
t0 *= tmp; t0 *= tmp;
t1 *= tmp; t1 *= tmp;
@ -14466,7 +14466,7 @@ void VPCALL idSIMD_SSE::DeriveTangents( idPlane *planes, idDrawVert *verts, cons
t5 = d3 * d7 - d2 * d8; t5 = d3 * d7 - d2 * d8;
tmp = idMath::RSqrt( t3 * t3 + t4 * t4 + t5 * t5 ); tmp = idMath::RSqrt( t3 * t3 + t4 * t4 + t5 * t5 );
*(unsigned long *)&tmp ^= signBit[0]; *(unsigned int *)&tmp ^= signBit[0];
t3 *= tmp; t3 *= tmp;
t4 *= tmp; t4 *= tmp;

8
idlib/math/Simd_SSE2.cpp
View file

@ -261,10 +261,10 @@ void VPCALL idSIMD_SSE2::CmpLT( byte *dst, const byte bitNum, const float *src0,
ALIGN8_INIT1( unsigned short SIMD_W_zero, 0 ); ALIGN8_INIT1( unsigned short SIMD_W_zero, 0 );
ALIGN8_INIT1( unsigned short SIMD_W_maxShort, 1<<15 ); ALIGN8_INIT1( unsigned short SIMD_W_maxShort, 1<<15 );
ALIGN4_INIT4( unsigned long SIMD_SP_singleSignBitMask, (unsigned long) ( 1 << 31 ), 0, 0, 0 ); ALIGN4_INIT4( unsigned int SIMD_SP_singleSignBitMask, (unsigned int) ( 1 << 31 ), 0, 0, 0 );
ALIGN4_INIT1( unsigned long SIMD_SP_signBitMask, (unsigned long) ( 1 << 31 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_signBitMask, (unsigned int) ( 1 << 31 ) );
ALIGN4_INIT1( unsigned long SIMD_SP_absMask, (unsigned long) ~( 1 << 31 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_absMask, (unsigned int) ~( 1 << 31 ) );
ALIGN4_INIT1( unsigned long SIMD_SP_infinityMask, (unsigned long) ~( 1 << 23 ) ); ALIGN4_INIT1( unsigned int SIMD_SP_infinityMask, (unsigned int) ~( 1 << 23 ) );
ALIGN4_INIT1( float SIMD_SP_zero, 0.0f ); ALIGN4_INIT1( float SIMD_SP_zero, 0.0f );
ALIGN4_INIT1( float SIMD_SP_one, 1.0f ); ALIGN4_INIT1( float SIMD_SP_one, 1.0f );