mirror of
https://github.com/alliedmodders/hl2sdk.git
synced 2025-09-20 20:46:03 +08:00
TF2 win64 + Ambuild tier1/mathlib + long=devil (#198)
* Fix compilation for windows, setup ambuild * Add built tier1 and mathlib for win64 * Ensure compilation is working on windows and linux * Add -fPIC * Add compiled libs, with optimisation enabled * Added windows lib * Fix hl2sdk for windows * Longs are the devil * Fix up threadtools functions * Add updated libs * Rework lib naming, and package script * Update lib directory according to new packaging --------- Co-authored-by: Kenzzer <kenzzer@users.noreply.github.com>
This commit is contained in:
Benoist
@ -16,8 +16,10 @@
|
||||
// memdbgon must be the last include file in a .cpp file!!!
|
||||
#include "tier0/memdbgon.h"
|
||||
|
||||
#if defined ( _WIN32 ) && !defined ( _WIN64 )
|
||||
static const uint32 _sincos_masks[] = { (uint32)0x0, (uint32)~0x0 };
|
||||
static const uint32 _sincos_inv_masks[] = { (uint32)~0x0, (uint32)0x0 };
|
||||
#endif
|
||||
|
||||
//-----------------------------------------------------------------------------
|
||||
// Macros and constants required by some of the SSE assembly:
|
||||
@ -49,6 +51,7 @@ static const uint32 _sincos_inv_masks[] = { (uint32)~0x0, (uint32)0x0 };
|
||||
static const __attribute__((aligned(16))) float _ps_##Name[4] = { Val, Val, Val, Val }
|
||||
#endif
|
||||
|
||||
#if defined ( _WIN32 ) && !defined ( _WIN64 )
|
||||
_PS_EXTERN_CONST(am_0, 0.0f);
|
||||
_PS_EXTERN_CONST(am_1, 1.0f);
|
||||
_PS_EXTERN_CONST(am_m1, -1.0f);
|
||||
@ -59,7 +62,7 @@ _PS_EXTERN_CONST(am_pi_o_2, (float)(M_PI / 2.0));
|
||||
_PS_EXTERN_CONST(am_2_o_pi, (float)(2.0 / M_PI));
|
||||
_PS_EXTERN_CONST(am_pi_o_4, (float)(M_PI / 4.0));
|
||||
_PS_EXTERN_CONST(am_4_o_pi, (float)(4.0 / M_PI));
|
||||
_PS_EXTERN_CONST_TYPE(am_sign_mask, int32, 0x80000000);
|
||||
_PS_EXTERN_CONST_TYPE(am_sign_mask, int32, (int32)0x80000000);
|
||||
_PS_EXTERN_CONST_TYPE(am_inv_sign_mask, int32, ~0x80000000);
|
||||
_PS_EXTERN_CONST_TYPE(am_min_norm_pos,int32, 0x00800000);
|
||||
_PS_EXTERN_CONST_TYPE(am_mant_mask, int32, 0x7f800000);
|
||||
@ -72,6 +75,7 @@ _PS_CONST(sincos_p0, 0.15707963267948963959e1f);
|
||||
_PS_CONST(sincos_p1, -0.64596409750621907082e0f);
|
||||
_PS_CONST(sincos_p2, 0.7969262624561800806e-1f);
|
||||
_PS_CONST(sincos_p3, -0.468175413106023168e-2f);
|
||||
#endif
|
||||
|
||||
#ifdef PFN_VECTORMA
|
||||
void __cdecl _SSE_VectorMA( const float *start, float scale, const float *direction, float *dest );
|
||||
@ -82,6 +86,9 @@ void __cdecl _SSE_VectorMA( const float *start, float scale, const float *direc
|
||||
//-----------------------------------------------------------------------------
|
||||
float _SSE_Sqrt(float x)
|
||||
{
|
||||
#if defined( _WIN64 )
|
||||
return std::sqrt(x);
|
||||
#else
|
||||
Assert( s_bMathlibInitialized );
|
||||
float root = 0.f;
|
||||
#ifdef _WIN32
|
||||
@ -100,6 +107,7 @@ float _SSE_Sqrt(float x)
|
||||
);
|
||||
#endif
|
||||
return root;
|
||||
#endif // _WIN64
|
||||
}
|
||||
|
||||
// Single iteration NewtonRaphson reciprocal square root:
|
||||
@ -123,6 +131,9 @@ float _SSE_RSqrtAccurate(float x)
|
||||
// Intel / Kipps SSE RSqrt. Significantly faster than above.
|
||||
float _SSE_RSqrtAccurate(float a)
|
||||
{
|
||||
#if defined( _WIN64 )
|
||||
return std::sqrt(a);
|
||||
#else
|
||||
float x;
|
||||
float half = 0.5f;
|
||||
float three = 3.f;
|
||||
@ -163,6 +174,7 @@ float _SSE_RSqrtAccurate(float a)
|
||||
#endif
|
||||
|
||||
return x;
|
||||
#endif // _WIN64
|
||||
}
|
||||
#endif
|
||||
|
||||
@ -170,6 +182,9 @@ float _SSE_RSqrtAccurate(float a)
|
||||
// or so, so ok for closed transforms. (ie, computing lighting normals)
|
||||
float _SSE_RSqrtFast(float x)
|
||||
{
|
||||
#if defined( _WIN64 )
|
||||
return std::sqrt(x);
|
||||
#else
|
||||
Assert( s_bMathlibInitialized );
|
||||
|
||||
float rroot = 0.0f;
|
||||
@ -192,10 +207,18 @@ float _SSE_RSqrtFast(float x)
|
||||
#endif
|
||||
|
||||
return rroot;
|
||||
#endif // _WIN64
|
||||
}
|
||||
|
||||
float FASTCALL _SSE_VectorNormalize (Vector& vec)
|
||||
{
|
||||
#if defined( _WIN64 )
|
||||
float l = std::sqrt(vec.x * vec.x + vec.y * vec.y + vec.z * vec.z);
|
||||
vec.x /= l;
|
||||
vec.y /= l;
|
||||
vec.z /= l;
|
||||
return l;
|
||||
#else
|
||||
Assert( s_bMathlibInitialized );
|
||||
|
||||
// NOTE: This is necessary to prevent an memory overwrite...
|
||||
@ -273,6 +296,7 @@ float FASTCALL _SSE_VectorNormalize (Vector& vec)
|
||||
}
|
||||
|
||||
return radius;
|
||||
#endif // _WIN64
|
||||
}
|
||||
|
||||
void FASTCALL _SSE_VectorNormalizeFast (Vector& vec)
|
||||
@ -286,6 +310,10 @@ void FASTCALL _SSE_VectorNormalizeFast (Vector& vec)
|
||||
|
||||
float _SSE_InvRSquared(const float* v)
|
||||
{
|
||||
#if defined( _WIN64 )
|
||||
float r2 = DotProduct(v, v);
|
||||
return r2 < 1.f ? 1.f : 1/r2;
|
||||
#else
|
||||
float inv_r2 = 1.f;
|
||||
#ifdef _WIN32
|
||||
_asm { // Intel SSE only routine
|
||||
@ -332,11 +360,15 @@ float _SSE_InvRSquared(const float* v)
|
||||
#endif
|
||||
|
||||
return inv_r2;
|
||||
#endif // _WIN64
|
||||
}
|
||||
|
||||
void _SSE_SinCos(float x, float* s, float* c)
|
||||
{
|
||||
#ifdef _WIN32
|
||||
#if defined( _WIN64 )
|
||||
*s = std::sin(x);
|
||||
*c = std::cos(x);
|
||||
#elif defined( _WIN32 )
|
||||
float t4, t8, t12;
|
||||
|
||||
__asm
|
||||
@ -430,7 +462,9 @@ void _SSE_SinCos(float x, float* s, float* c)
|
||||
|
||||
float _SSE_cos( float x )
|
||||
{
|
||||
#ifdef _WIN32
|
||||
#if defined ( _WIN64 )
|
||||
return std::cos(x);
|
||||
#elif defined( _WIN32 )
|
||||
float temp;
|
||||
__asm
|
||||
{
|
||||
@ -493,7 +527,10 @@ float _SSE_cos( float x )
|
||||
//-----------------------------------------------------------------------------
|
||||
void _SSE2_SinCos(float x, float* s, float* c) // any x
|
||||
{
|
||||
#ifdef _WIN32
|
||||
#if defined( _WIN64 )
|
||||
*s = std::sin(x);
|
||||
*c = std::cos(x);
|
||||
#elif defined( _WIN32 )
|
||||
__asm
|
||||
{
|
||||
movss xmm0, x
|
||||
@ -578,7 +615,9 @@ void _SSE2_SinCos(float x, float* s, float* c) // any x
|
||||
|
||||
float _SSE2_cos(float x)
|
||||
{
|
||||
#ifdef _WIN32
|
||||
#if defined ( _WIN64 )
|
||||
return std::cos(x);
|
||||
#elif defined( _WIN32 )
|
||||
__asm
|
||||
{
|
||||
movss xmm0, x
|
||||
@ -638,8 +677,11 @@ void VectorTransformSSE(const float *in1, const matrix3x4_t& in2, float *out1)
|
||||
{
|
||||
Assert( s_bMathlibInitialized );
|
||||
Assert( in1 != out1 );
|
||||
|
||||
#ifdef _WIN32
|
||||
#if defined ( _WIN64 )
|
||||
out1[0] = DotProduct(in1, in2[0]) + in2[0][3];
|
||||
out1[1] = DotProduct(in1, in2[1]) + in2[1][3];
|
||||
out1[2] = DotProduct(in1, in2[2]) + in2[2][3];
|
||||
#elif defined( _WIN32 )
|
||||
__asm
|
||||
{
|
||||
mov eax, in1;
|
||||
@ -695,8 +737,11 @@ void VectorRotateSSE( const float *in1, const matrix3x4_t& in2, float *out1 )
|
||||
{
|
||||
Assert( s_bMathlibInitialized );
|
||||
Assert( in1 != out1 );
|
||||
|
||||
#ifdef _WIN32
|
||||
#if defined ( _WIN64 )
|
||||
out1[0] = DotProduct( in1, in2[0] );
|
||||
out1[1] = DotProduct( in1, in2[1] );
|
||||
out1[2] = DotProduct( in1, in2[2] );
|
||||
#elif defined( _WIN32 )
|
||||
__asm
|
||||
{
|
||||
mov eax, in1;
|
||||
@ -745,7 +790,7 @@ void VectorRotateSSE( const float *in1, const matrix3x4_t& in2, float *out1 )
|
||||
#endif
|
||||
}
|
||||
|
||||
#ifdef _WIN32
|
||||
#if defined( _WIN32 ) && !defined( _WIN64 )
|
||||
void _declspec(naked) _SSE_VectorMA( const float *start, float scale, const float *direction, float *dest )
|
||||
{
|
||||
// FIXME: This don't work!! It will overwrite memory in the write to dest
|
||||
@ -776,7 +821,7 @@ void _declspec(naked) _SSE_VectorMA( const float *start, float scale, const floa
|
||||
}
|
||||
#endif
|
||||
|
||||
#ifdef _WIN32
|
||||
#if defined( _WIN32 ) && !defined( _WIN64 )
|
||||
#ifdef PFN_VECTORMA
|
||||
void _declspec(naked) __cdecl _SSE_VectorMA( const Vector &start, float scale, const Vector &direction, Vector &dest )
|
||||
{
|
||||
@ -841,5 +886,4 @@ vec_t DotProduct (const vec_t *a, const vec_t *c)
|
||||
ret
|
||||
}
|
||||
}
|
||||
*/
|
||||
|
||||
*/
|
||||
Reference in New Issue
Block a user