mirror of
https://github.com/ptitSeb/Serious-Engine
synced 2024-11-29 21:25:54 +01:00
df6199de70
Sdl2 port
612 lines
13 KiB
C++
Executable File
612 lines
13 KiB
C++
Executable File
/* Copyright (c) 2002-2012 Croteam Ltd.
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of version 2 of the GNU General Public License as published by
|
|
the Free Software Foundation
|
|
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License along
|
|
with this program; if not, write to the Free Software Foundation, Inc.,
|
|
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. */
|
|
|
|
#ifndef SE_INCL_FUNCTIONS_H
|
|
#define SE_INCL_FUNCTIONS_H
|
|
#ifdef PRAGMA_ONCE
|
|
#pragma once
|
|
#endif
|
|
|
|
// asm shortcuts
|
|
#define O offset
|
|
#define Q qword ptr
|
|
#define D dword ptr
|
|
#define W word ptr
|
|
#define B byte ptr
|
|
|
|
/*
|
|
* template implementations
|
|
*/
|
|
|
|
template<class Type>
|
|
inline Type Abs( const Type x)
|
|
{
|
|
return ( x>=Type(0) ? x : -x );
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type Max( const Type a, const Type b)
|
|
{
|
|
return ( a<b ? b : a );
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type Min( const Type a, const Type b)
|
|
{
|
|
return ( a>b ? b : a );
|
|
}
|
|
|
|
// linear interpolation
|
|
template<class Type>
|
|
inline Type Lerp( const Type x0, const Type x1, const FLOAT fRatio)
|
|
{
|
|
if( fRatio==0) return x0;
|
|
else if( fRatio==1) return x1;
|
|
else return ((Type) (x0+(x1-x0)*fRatio));
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type Sgn( const Type x)
|
|
{
|
|
return (x)>Type(0) ? Type(1):( x<0 ? Type(-1):Type(0) );
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type SgnNZ( const Type x)
|
|
{
|
|
return (x)>=Type(0) ? Type(1):Type(-1);
|
|
}
|
|
|
|
template<class Type>
|
|
inline void Swap( Type &a, Type &b)
|
|
{
|
|
Type t=a; a=b; b=t;
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type ClampUp( const Type x, const Type uplimit)
|
|
{
|
|
return ( x<=uplimit ? x : uplimit );
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type ClampDn( const Type x, const Type dnlimit)
|
|
{
|
|
return ( x>=dnlimit ? x : dnlimit );
|
|
}
|
|
|
|
template<class Type>
|
|
inline Type Clamp( const Type x, const Type dnlimit, const Type uplimit)
|
|
{
|
|
return ( x>=dnlimit ? (x<=uplimit ? x : uplimit): dnlimit );
|
|
}
|
|
|
|
/*
|
|
* fast implementations
|
|
*/
|
|
|
|
|
|
inline DOUBLE Abs( const DOUBLE f) { return fabs(f); }
|
|
inline FLOAT Abs( const FLOAT f) { return (FLOAT)fabs(f); }
|
|
inline SLONG Abs( const SLONG sl) { return labs(sl); }
|
|
|
|
|
|
/*
|
|
inline FLOAT Min( const FLOAT fA, const FLOAT fB)
|
|
{
|
|
FLOAT fRet;
|
|
__asm {
|
|
fld D [fA]
|
|
fld D [fB]
|
|
fucomi st(0),st(1)
|
|
fcmovnb st(0),st(1)
|
|
ffree st(1)
|
|
fstp D [fRet]
|
|
}
|
|
return fRet;
|
|
}
|
|
|
|
inline FLOAT Max( const FLOAT fA, const FLOAT fB)
|
|
{
|
|
FLOAT fRet;
|
|
__asm {
|
|
fld D [fA]
|
|
fld D [fB]
|
|
fucomi st(0),st(1)
|
|
fcmovb st(0),st(1)
|
|
ffree st(1)
|
|
fstp D [fRet]
|
|
}
|
|
return fRet;
|
|
}
|
|
|
|
|
|
inline SLONG Min( const SLONG slA, const SLONG slB)
|
|
{
|
|
SLONG slRet;
|
|
__asm {
|
|
mov eax,D [slA]
|
|
cmp eax,D [slB]
|
|
cmovg eax,D [slB]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
}
|
|
|
|
inline ULONG Min( const ULONG slA, const ULONG slB)
|
|
{
|
|
ULONG ulRet;
|
|
__asm {
|
|
mov eax,D [slA]
|
|
cmp eax,D [slB]
|
|
cmova eax,D [slB]
|
|
mov D [ulRet],eax
|
|
}
|
|
return ulRet;
|
|
}
|
|
|
|
inline SLONG Max( const SLONG slA, const SLONG slB)
|
|
{
|
|
SLONG slRet;
|
|
__asm {
|
|
mov eax,D [slA]
|
|
cmp eax,D [slB]
|
|
cmovl eax,D [slB]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
}
|
|
|
|
inline ULONG Max( const ULONG slA, const ULONG slB)
|
|
{
|
|
ULONG ulRet;
|
|
__asm {
|
|
mov eax,D [slA]
|
|
cmp eax,D [slB]
|
|
cmovb eax,D [slB]
|
|
mov D [ulRet],eax
|
|
}
|
|
return ulRet;
|
|
}
|
|
|
|
|
|
|
|
inline FLOAT ClampUp( const FLOAT f, const FLOAT fuplimit)
|
|
{
|
|
FLOAT fRet;
|
|
__asm {
|
|
fld D [fuplimit]
|
|
fld D [f]
|
|
fucomi st(0),st(1)
|
|
fcmovnb st(0),st(1)
|
|
fstp D [fRet]
|
|
fstp st(0)
|
|
}
|
|
return fRet;
|
|
}
|
|
|
|
inline FLOAT ClampDn( const FLOAT f, const FLOAT fdnlimit)
|
|
{
|
|
FLOAT fRet;
|
|
__asm {
|
|
fld D [fdnlimit]
|
|
fld D [f]
|
|
fucomi st(0),st(1)
|
|
fcmovb st(0),st(1)
|
|
fstp D [fRet]
|
|
fstp st(0)
|
|
}
|
|
return fRet;
|
|
}
|
|
|
|
inline FLOAT Clamp( const FLOAT f, const FLOAT fdnlimit, const FLOAT fuplimit)
|
|
{
|
|
FLOAT fRet;
|
|
__asm {
|
|
fld D [fdnlimit]
|
|
fld D [fuplimit]
|
|
fld D [f]
|
|
fucomi st(0),st(2)
|
|
fcmovb st(0),st(2)
|
|
fucomi st(0),st(1)
|
|
fcmovnb st(0),st(1)
|
|
fstp D [fRet]
|
|
fcompp
|
|
}
|
|
return fRet;
|
|
}
|
|
|
|
|
|
inline SLONG ClampDn( const SLONG sl, const SLONG sldnlimit)
|
|
{
|
|
SLONG slRet;
|
|
__asm {
|
|
mov eax,D [sl]
|
|
cmp eax,D [sldnlimit]
|
|
cmovl eax,D [sldnlimit]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
}
|
|
|
|
inline SLONG ClampUp( const SLONG sl, const SLONG sluplimit)
|
|
{
|
|
SLONG slRet;
|
|
__asm {
|
|
mov eax,D [sl]
|
|
cmp eax,D [sluplimit]
|
|
cmovg eax,D [sluplimit]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
}
|
|
|
|
inline SLONG Clamp( const SLONG sl, const SLONG sldnlimit, const SLONG sluplimit)
|
|
{
|
|
SLONG slRet;
|
|
__asm {
|
|
mov eax,D [sl]
|
|
cmp eax,D [sldnlimit]
|
|
cmovl eax,D [sldnlimit]
|
|
cmp eax,D [sluplimit]
|
|
cmovg eax,D [sluplimit]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
}
|
|
|
|
*/
|
|
|
|
/*
|
|
* fast functions
|
|
*/
|
|
|
|
#define FP_ONE_BITS 0x3F800000
|
|
|
|
// fast reciprocal value
|
|
inline FLOAT FastRcp( const FLOAT f)
|
|
{
|
|
INDEX i = 2*FP_ONE_BITS - *(INDEX*)&(f);
|
|
FLOAT r = *(FLOAT*)&i;
|
|
return( r * (2.0f - f*r));
|
|
}
|
|
|
|
// convert float from 0.0f to 1.0f -> ulong form 0 to 255
|
|
inline ULONG NormFloatToByte( const FLOAT f)
|
|
{
|
|
/* rcg10042001 !!! FIXME: Move this elsewhere. */
|
|
#ifdef __MSVC_INLINE__
|
|
const FLOAT f255 = 255.0f;
|
|
ULONG ulRet;
|
|
__asm {
|
|
fld D [f]
|
|
fmul D [f255]
|
|
fistp D [ulRet]
|
|
}
|
|
return ulRet;
|
|
#else
|
|
ASSERT((f >= 0.0f) && (f <= 1.0f));
|
|
return( (ULONG) (f * 255.0f) );
|
|
#endif
|
|
}
|
|
|
|
// convert ulong from 0 to 255 -> float form 0.0f to 255.0f
|
|
inline FLOAT NormByteToFloat( const ULONG ul)
|
|
{
|
|
return (FLOAT)ul * (1.0f/255.0f);
|
|
}
|
|
|
|
|
|
// fast float to int conversion
|
|
inline SLONG FloatToInt( FLOAT f)
|
|
{
|
|
#if defined(__arm__) || defined(USE_PORTABLE_C)
|
|
return((SLONG) (f + 0.5f)); /* best of luck to you. */
|
|
|
|
#elif (defined __MSVC_INLINE__)
|
|
SLONG slRet;
|
|
__asm {
|
|
fld D [f]
|
|
fistp D [slRet]
|
|
}
|
|
return slRet;
|
|
|
|
#elif (defined __GNU_INLINE__)
|
|
SLONG slRet;
|
|
__asm__ __volatile__ (
|
|
"flds (%%eax) \n\t"
|
|
"fistpl (%%esi) \n\t"
|
|
:
|
|
: "a" (&f), "S" (&slRet)
|
|
: "memory"
|
|
);
|
|
return(slRet);
|
|
#else
|
|
#error Fill this in for your platform.
|
|
#endif
|
|
}
|
|
|
|
// log base 2 of any float numero
|
|
inline FLOAT Log2( FLOAT f) {
|
|
#if (defined USE_PORTABLE_C) || defined(__arm__)
|
|
// !!! FIXME: What's wrong with log2()?
|
|
return (FLOAT)(log10(f)*3.321928094887); // log10(x)/log10(2)
|
|
|
|
#elif (defined __MSVC_INLINE__)
|
|
FLOAT fRet;
|
|
_asm {
|
|
fld1
|
|
fld D [f]
|
|
fyl2x
|
|
fstp D [fRet]
|
|
}
|
|
return fRet;
|
|
|
|
#elif (defined __GNU_INLINE__)
|
|
FLOAT fRet;
|
|
__asm__ __volatile__ (
|
|
"fld1 \n\t"
|
|
"flds (%%eax) \n\t"
|
|
"fyl2x \n\t"
|
|
"fstps (%%esi) \n\t"
|
|
:
|
|
: "a" (&f), "S" (&fRet)
|
|
: "memory"
|
|
);
|
|
return(fRet);
|
|
#else
|
|
#error Fill this in for your platform.
|
|
#endif
|
|
}
|
|
|
|
|
|
// returns accurate values only for integers that are power of 2
|
|
inline SLONG FastLog2( SLONG x)
|
|
{
|
|
#if (defined USE_PORTABLE_C)
|
|
register SLONG val = x;
|
|
register SLONG retval = 31;
|
|
while (retval > 0)
|
|
{
|
|
if (val & (1 << retval))
|
|
return retval;
|
|
retval--;
|
|
}
|
|
|
|
return 0;
|
|
|
|
#elif (defined __MSVC_INLINE__)
|
|
SLONG slRet;
|
|
__asm {
|
|
bsr eax,D [x]
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
|
|
#elif (defined __GNU_INLINE__)
|
|
SLONG slRet;
|
|
__asm__ __volatile__ (
|
|
"bsrl %%ecx, %%eax \n\t"
|
|
: "=a" (slRet)
|
|
: "c" (x)
|
|
: "memory"
|
|
);
|
|
return(slRet);
|
|
#else
|
|
#error Fill this in for your platform.
|
|
#endif
|
|
}
|
|
|
|
// returns log2 of first larger value that is a power of 2
|
|
inline SLONG FastMaxLog2( SLONG x)
|
|
{
|
|
#if (defined USE_PORTABLE_C)
|
|
printf("CHECK THIS: %s:%d\n", __FILE__, __LINE__);
|
|
return((SLONG) log2((double) x));
|
|
|
|
#elif (defined __MSVC_INLINE__)
|
|
SLONG slRet;
|
|
__asm {
|
|
bsr eax,D [x]
|
|
bsf edx,D [x]
|
|
cmp edx,eax
|
|
adc eax,0
|
|
mov D [slRet],eax
|
|
}
|
|
return slRet;
|
|
|
|
#elif (defined __GNU_INLINE__)
|
|
SLONG slRet;
|
|
__asm__ __volatile__ (
|
|
"bsrl %%ecx, %%eax \n\t"
|
|
"bsfl %%ecx, %%edx \n\t"
|
|
"cmpl %%eax, %%edx \n\t"
|
|
"adcl $0, %%eax \n\t"
|
|
: "=a" (slRet)
|
|
: "c" (x)
|
|
: "memory"
|
|
);
|
|
return(slRet);
|
|
#else
|
|
#error Fill this in for your platform.
|
|
#endif
|
|
}
|
|
|
|
|
|
|
|
// square root (works with negative numbers)
|
|
#ifdef __arm__
|
|
inline FLOAT Sqrt( FLOAT x) { return sqrtf( ClampDn( x, 0.0f)); }
|
|
#else
|
|
inline FLOAT Sqrt( FLOAT x) { return (FLOAT)sqrt( ClampDn( x, 0.0f)); }
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
* Trigonometrical functions
|
|
*/
|
|
|
|
//#define ANGLE_MASK 0x3fff
|
|
#define ANGLE_SNAP (0.25f) //0x0010
|
|
// Wrap angle to be between 0 and 360 degrees
|
|
inline ANGLE WrapAngle(ANGLE a) {
|
|
return (ANGLE) fmod( fmod(a,360.0) + 360.0, 360.0); // 0..360
|
|
}
|
|
|
|
// Normalize angle to be between -180 and +180 degrees
|
|
inline ANGLE NormalizeAngle(ANGLE a) {
|
|
return WrapAngle(a+ANGLE_180)-ANGLE_180;
|
|
}
|
|
|
|
// math constants
|
|
static const FLOAT PI = FLOAT(3.14159265359);
|
|
|
|
// convert degrees into angle
|
|
inline ANGLE AngleDeg(FLOAT fDegrees) {
|
|
//return ANGLE (fDegrees*ANGLE_180/FLOAT(180.0));
|
|
return fDegrees;
|
|
}
|
|
// convert radians into angle
|
|
inline ANGLE AngleRad(FLOAT fRadians) {
|
|
return ANGLE (fRadians*ANGLE_180/PI);
|
|
}
|
|
// convert radians into angle
|
|
inline ANGLE AngleRad(DOUBLE dRadians) {
|
|
return ANGLE (dRadians*ANGLE_180/PI);
|
|
}
|
|
// convert angle into degrees
|
|
inline FLOAT DegAngle(ANGLE aAngle) {
|
|
//return FLOAT (WrapAngle(aAngle)*FLOAT(180.0)/ANGLE_180);
|
|
return WrapAngle(aAngle);
|
|
}
|
|
// convert angle into radians
|
|
inline FLOAT RadAngle(ANGLE aAngle) {
|
|
return FLOAT (WrapAngle(aAngle)*PI/ANGLE_180);
|
|
}
|
|
|
|
ENGINE_API FLOAT Sin(ANGLE a);
|
|
ENGINE_API FLOAT Cos(ANGLE a);
|
|
ENGINE_API FLOAT Tan(ANGLE a);
|
|
|
|
#ifdef __arm__
|
|
inline ENGINE_API FLOAT SinFast(ANGLE a) { return sinf(a*(PI/ANGLE_180)); };
|
|
inline ENGINE_API FLOAT CosFast(ANGLE a) { return cosf(a*(PI/ANGLE_180)); };
|
|
inline ENGINE_API FLOAT TanFast(ANGLE a) { return tanf(a*(PI/ANGLE_180)); };
|
|
inline ANGLE ASin(FLOAT y) {
|
|
return AngleRad (asinf(Clamp(y, -1.0f, 1.0f)));
|
|
}
|
|
inline ANGLE ACos(FLOAT x) {
|
|
return AngleRad (acosf(Clamp(x, -1.0f, 1.0f)));
|
|
|
|
}
|
|
inline ANGLE ATan(FLOAT z) {
|
|
return AngleRad (atanf(z));
|
|
}
|
|
inline ANGLE ATan2(FLOAT y, FLOAT x) {
|
|
return AngleRad (atan2f(y, x));
|
|
}
|
|
#else
|
|
inline ENGINE_API FLOAT SinFast(ANGLE a) { return (FLOAT)sin(a*(PI/ANGLE_180)); };
|
|
inline ENGINE_API FLOAT CosFast(ANGLE a) { return (FLOAT)cos(a*(PI/ANGLE_180)); };
|
|
inline ENGINE_API FLOAT TanFast(ANGLE a) { return (FLOAT)tan(a*(PI/ANGLE_180)); };
|
|
inline ANGLE ASin(FLOAT y) {
|
|
return AngleRad (asin(Clamp(y, -1.0f, 1.0f)));
|
|
}
|
|
inline ANGLE ACos(FLOAT x) {
|
|
return AngleRad (acos(Clamp(x, -1.0f, 1.0f)));
|
|
}
|
|
inline ANGLE ATan(FLOAT z) {
|
|
return AngleRad (atan(z));
|
|
}
|
|
inline ANGLE ATan2(FLOAT y, FLOAT x) {
|
|
return AngleRad (atan2(y, x));
|
|
}
|
|
#endif
|
|
|
|
inline ANGLE ASin(DOUBLE y) {
|
|
return AngleRad (asin(Clamp(y, -1.0, 1.0)));
|
|
}
|
|
inline ANGLE ACos(DOUBLE x) {
|
|
return AngleRad (acos(Clamp(x, -1.0, 1.0)));
|
|
}
|
|
inline ANGLE ATan(DOUBLE z) {
|
|
return AngleRad (atan(z));
|
|
}
|
|
inline ANGLE ATan2(DOUBLE y, DOUBLE x) {
|
|
return AngleRad (atan2(y, x));
|
|
}
|
|
|
|
// does "snap to grid" for given coordinate
|
|
ENGINE_API void Snap( FLOAT &fDest, FLOAT fStep);
|
|
ENGINE_API void Snap( DOUBLE &fDest, DOUBLE fStep);
|
|
// does "snap to grid" for given angle
|
|
//ENGINE_API void Snap( ANGLE &angDest, ANGLE angStep);
|
|
|
|
|
|
/*
|
|
* linear interpolation, special functions for floats and angles
|
|
*/
|
|
|
|
inline FLOAT LerpFLOAT(FLOAT f0, FLOAT f1, FLOAT fFactor)
|
|
{
|
|
return f0+(f1-f0)*fFactor;
|
|
}
|
|
|
|
inline ANGLE LerpANGLE(ANGLE a0, ANGLE a1, FLOAT fFactor)
|
|
{
|
|
// calculate delta
|
|
ANGLE aDelta = WrapAngle(a1)-WrapAngle(a0);
|
|
// adjust delta not to wrap around 360
|
|
if (aDelta>ANGLE_180) {
|
|
aDelta-=ANGLE(ANGLE_360);
|
|
} else if (aDelta<-ANGLE_180) {
|
|
aDelta+=ANGLE(ANGLE_360);
|
|
}
|
|
// interpolate the delta
|
|
return a0+ANGLE(fFactor*aDelta);
|
|
}
|
|
|
|
// Calculates ratio function /~~\ where 0<x<1, taking in consideration fade in and fade out percentages
|
|
// (ie. 0.2f means 20% fade in, 0.1f stands for 10% fade out)
|
|
inline FLOAT CalculateRatio(FLOAT fCurr, FLOAT fMin, FLOAT fMax, FLOAT fFadeInRatio, FLOAT fFadeOutRatio)
|
|
{
|
|
if(fCurr<=fMin || fCurr>=fMax)
|
|
{
|
|
return 0.0f;
|
|
}
|
|
FLOAT fDelta = fMax-fMin;
|
|
FLOAT fRatio=(fCurr-fMin)/fDelta;
|
|
if(fRatio<fFadeInRatio) {
|
|
fRatio = Clamp( fRatio/fFadeInRatio, 0.0f, 1.0f);
|
|
} else if(fRatio>(1-fFadeOutRatio)) {
|
|
fRatio = Clamp( (1.0f-fRatio)/fFadeOutRatio, 0.0f, 1.0f);
|
|
} else {
|
|
fRatio = 1.0f;
|
|
}
|
|
return fRatio;
|
|
}
|
|
|
|
|
|
#undef O
|
|
#undef Q
|
|
#undef D
|
|
#undef W
|
|
#undef B
|
|
|
|
#endif /* include-once check. */
|
|
|
|
|