/*========================== begin_copyright_notice ============================

Copyright (C) 2017-2021 Intel Corporation

SPDX-License-Identifier: MIT

============================= end_copyright_notice ===========================*/

#include "spirv.h"
#include "IMF/FP32/asin_s_la.cl"

INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fclamp, _f32_f32_f32, )(float x, float minval, float maxval ){
    return SPIRV_OCL_BUILTIN(fmin, _f32_f32, )(SPIRV_OCL_BUILTIN(fmax, _f32_f32, )(x, minval), maxval);
}

INLINE
uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_abs, _i8, )(char x ){
    return x > 0 ? x : -x;
}

INLINE
ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_abs, _i16, )(short x ){
    return x > 0 ? x : -x;
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_abs, _i32, )(int x ){
    // Convert signed to unsigned to have consistent result.
    // Also, make abs(INT32_MIN) well defined : abs(INT32_MIN) = (ui32)(-(i64)INT32_MIN)
    uint ux = (uint)x;
    return x > 0 ? ux : -ux;
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_add_sat, _i8_i8, )( char x,
                                      char y )
{
    return (char)SPIRV_OCL_BUILTIN(s_min, _i32_i32, )( SPIRV_OCL_BUILTIN(s_max, _i32_i32, )( (int)x + (int)y , CHAR_MIN), CHAR_MAX);
}

INLINE
uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_add_sat, _i8_i8, )( uchar x,
                                       uchar y )
{
    return (uchar)SPIRV_OCL_BUILTIN(u_min, _i32_i32, )( SPIRV_OCL_BUILTIN(u_max, _i32_i32, )( (uint)x + (uint)y , 0), UCHAR_MAX);
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_add_sat, _i16_i16, )( short x,
                                         short y )
{
    return (short)SPIRV_OCL_BUILTIN(s_min, _i32_i32, )( SPIRV_OCL_BUILTIN(s_max, _i32_i32, )( (int)x + (int)y , SHRT_MIN), SHRT_MAX);
}

INLINE
ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_add_sat, _i16_i16, )( ushort x,
                                          ushort y )
{
    return (ushort)SPIRV_OCL_BUILTIN(u_min, _i32_i32, )(SPIRV_OCL_BUILTIN(u_max, _i32_i32, )( (uint)x + (uint)y, 0), USHRT_MAX);
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_add_sat, _i32_i32, )( int x,
                                       int y )
{
    long tmp = (long) x + (long) y;
    return (int) SPIRV_OCL_BUILTIN(s_clamp, _i64_i64_i64, )( tmp, (long)INT_MIN, (long)INT_MAX);
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_add_sat, _i32_i32, )( uint x,
                                        uint y )
{
    long tmp = (long) x + (long) y;
    return (uint) SPIRV_OCL_BUILTIN(s_clamp, _i64_i64_i64, )( tmp, 0L, (long)UINT_MAX);
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ctz, _i8, )(char x ){
    return (char)SPIRV_OCL_BUILTIN(ctz, _i32, )(as_int((uint)x | (1U << sizeof(x) * 8)));
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ctz, _i16, )(short x ){
    return (short)SPIRV_OCL_BUILTIN(ctz, _i32, )(as_int((uint)x | (1U << sizeof(x) * 8)));
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ctz, _i32, )(int x ){
    uint rev = __builtin_IB_bfrev(x);
    return SPIRV_OCL_BUILTIN(clz, _i32, )(as_int(rev));
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_mad24, _i32_i32_i32, )( int x,
                                         int y,
                                         int z )
{
    return SPIRV_OCL_BUILTIN(s_mul24, _i32_i32, )(x, y) + z;
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_mad24, _i32_i32_i32, )( uint x,
                                          uint y,
                                          uint z )
{
    return SPIRV_OCL_BUILTIN(u_mul24, _i32_i32, )(x, y) + z;
}

INLINE char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_mad_sat, _i8_i8_i8, )(
    char a,
    char b,
    char c)
{
    short res = (short)a * (short)b + (short)c;
    return SPIRV_BUILTIN(SConvert, _Sat_i8_i16, _Rchar_sat)(res);
}

INLINE uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_mad_sat, _i8_i8_i8, )(
    uchar a,
    uchar b,
    uchar c)
{
    ushort res = (ushort)a * (ushort)b + (ushort)c;
    return SPIRV_BUILTIN(UConvert, _Sat_i8_i16, _Ruchar_sat)(res);
}

INLINE short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_mad_sat, _i16_i16_i16, )(
    short a,
    short b,
    short c)
{
    int res = (int)a * (int)b + (int)c;
    return SPIRV_BUILTIN(SConvert, _Sat_i16_i32, _Rshort_sat)(res);
}

INLINE ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_mad_sat, _i16_i16_i16, )(
    ushort a,
    ushort b,
    ushort c)
{
    uint res = (uint)a * (uint)b + (uint)c;
    return SPIRV_BUILTIN(UConvert, _Sat_i16_i32, _Rushort_sat)(res);
}

INLINE int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_mad_sat, _i32_i32_i32, )(
    int a,
    int b,
    int c)
{
    long res = (long)a * (long)b + (long)c;
    return SPIRV_BUILTIN(SConvert, _Sat_i32_i64, _Rint_sat)(res);
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_mad_sat, _i32_i32_i32, )( uint a,
                                            uint b,
                                            uint c )
{
    ulong res = (ulong)a * (ulong)b + (ulong)c;
    return SPIRV_BUILTIN(UConvert, _Sat_i32_i64, _Ruint_sat)(res);
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_max, _i8_i8, )(char x, char y ){
    return (x >= y) ? x : y;
}

INLINE
uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_max, _i8_i8, )(uchar x, uchar y ){
    return (x >= y) ? x : y;
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_max, _i16_i16, )(short x, short y ){
    return (x >= y) ? x : y;
}

INLINE
ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_max, _i16_i16, )(ushort x, ushort y ){
    return (x >= y) ? x : y;
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_max, _i32_i32, )(int x, int y ){
    return (x >= y) ? x : y;
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_max, _i32_i32, )(uint x, uint y ){
    return (x >= y) ? x : y;
}

INLINE
long SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_max, _i64_i64, )(long x, long y ){
    return (x >= y) ? x : y;
}

INLINE
ulong SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_max, _i64_i64, )(ulong x, ulong y ){
    return (x >= y) ? x : y;
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_min, _i8_i8, )(char x, char y ){
    return (x <= y) ? x : y;
}

INLINE
uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_min, _i8_i8, )(uchar x, uchar y ){
    return (x <= y) ? x : y;
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_min, _i16_i16, )(short x, short y ){
    return (x <= y) ? x : y;
}

INLINE
ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_min, _i16_i16, )(ushort x, ushort y ){
    return (x <= y) ? x : y;
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_min, _i32_i32, )(int x, int y ){
    return (x <= y) ? x : y;
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_min, _i32_i32, )(uint x, uint y ){
    return (x <= y) ? x : y;
}

INLINE
long SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_min, _i64_i64, )(long x, long y ){
    return (x < y) ? x : y;
}

INLINE
ulong SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_min, _i64_i64, )(ulong x, ulong y ){
    return (x < y) ? x : y;
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_mul_hi, _i32_i32, )( uint x,
                                       uint y )
{
    return ((ulong)x * (ulong)y) >> 32;
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_mul_hi, _i32_i32, )( int x,
                                      int y )
{
    return ((long)x * (long)y) >> 32;
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(popcount, _i8, )(char x ){
    return __builtin_IB_popcount_1u8(x);
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(popcount, _i16, )(short x ){
    return __builtin_IB_popcount_1u16(x);
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(popcount, _i32, )(int x ){
    return __builtin_IB_popcount_1u32(x);
}

INLINE
char SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_sub_sat, _i8_i8, )( char x,
                                      char y )
{
    short tmp = (short)x - (short)y;
    return (char) SPIRV_OCL_BUILTIN(s_clamp, _i16_i16_i16, )(tmp, (short)CHAR_MIN, (short)CHAR_MAX);
}

INLINE
uchar SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_sub_sat, _i8_i8, )( uchar x,
                                       uchar y )
{
    return ( x <= y ) ? 0 : x - y;
}

INLINE
short SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_sub_sat, _i16_i16, )( short x,
                                         short y )
{
    int tmp = (int)x - (int)y;
    return (short) SPIRV_OCL_BUILTIN(s_clamp, _i32_i32_i32, )( tmp, (int)SHRT_MIN, (int)SHRT_MAX);
}

INLINE
ushort SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_sub_sat, _i16_i16, )( ushort x,
                                          ushort y )
{
    return ( x <= y ) ? 0 : x - y;
}

INLINE
int SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(s_sub_sat, _i32_i32, )( int x,
                                       int y )
{
    long tmp = (long)x - (long)y;
    return (int) SPIRV_OCL_BUILTIN(s_clamp, _i64_i64_i64, )( tmp, (long)INT_MIN, (long)INT_MAX);
}

INLINE
uint SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(u_sub_sat, _i32_i32, )( uint x,
                                        uint y )
{
    return ( x <= y ) ? 0 : x - y;
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(acos, _f32, )(float x ){
    // The LA acos implementation (IMF/FP32/acos_s_la.cl)
    // seems to be slower on Mandelbulb algorithm..
    float temp1 = 0.0f;
    float temp2 = 0.0f;
    float temp4 = 0.0f;

    float destTemp = 0.0f;

    bool flag = false;

    temp1 = -SPIRV_OCL_BUILTIN(fabs, _f32, )(x) + 1.0f;

    temp1 = temp1  * 0.5f;

    flag = SPIRV_OCL_BUILTIN(fabs, _f32, )(x) > 0.575f;

    if( flag )
    {
        temp1 = SPIRV_OCL_BUILTIN(sqrt, _f32, )(temp1);
    }
    else
    {
        temp1 = SPIRV_OCL_BUILTIN(fabs, _f32, )(x);
    }

    temp2 = temp1 * temp1;

    destTemp = temp2 * -0.5011622905731201f;

    temp4 = temp2 + -5.478654384613037f;

    destTemp = destTemp + 0.9152014851570129f;

    temp4 = temp2 * temp4;

    destTemp = temp2 * destTemp;

    temp4 = temp4 + 5.491230487823486f;

    destTemp = temp1 * destTemp;

    temp4 = SPIRV_OCL_BUILTIN(native_recip, _f32, )(temp4);

    destTemp = temp4 * destTemp;

    destTemp = temp1 + destTemp;

    if( flag )
    {
        destTemp = destTemp * 2.0f;
    }
    else
    {
        destTemp = -destTemp + 1.5707963705062866f;
    }

    if( x <= 0.0 )
    {
        destTemp = -destTemp + 3.1415927410125732f;
    }

    return destTemp;
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(asin, _f32, )(float value ){
    return __ocl_svml_asinf(value);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(atan, _f32, )(float value ){
    // The LA atan implementation (IMF/FP32/atan_s_la.cl)
    // seems to be slower on Mandelbulb algorithm..
    float temp1 = 0.0f;
    float temp2 = 0.0f;
    float temp3 = 0.0f;
    float temp4 = 0.0f;

    float destTemp = 0.0f;

    bool flag = SPIRV_OCL_BUILTIN(fabs, _f32, )(value) > 1.0f;

    temp1 = SPIRV_OCL_BUILTIN(fabs, _f32, )(value);

    if(flag)
    {
        temp1 = SPIRV_OCL_BUILTIN(native_recip, _f32, )(temp1);
    }

    temp2 = temp1 * temp1;

    destTemp = temp2 * -0.8233629465103149f;

    temp4 = temp2 + 11.33538818359375f;

    destTemp = destTemp + -5.674867153167725f;

    temp4 = temp4 * temp2;

    destTemp = temp2 * destTemp;

    temp4 = temp4 + 28.84246826171875f;

    destTemp = destTemp + -6.565555095672607f;

    temp4 = temp4 * temp2;

    destTemp = temp2 * destTemp;

    temp4 = temp4 + 19.696670532226562f;

    destTemp = temp1 * destTemp;

    temp4 = SPIRV_OCL_BUILTIN(native_recip, _f32, )(temp4);

    destTemp = temp4 * destTemp;

    destTemp = destTemp + temp1;

    if(flag)
    {
        destTemp = -destTemp + 1.5707963705062866f;
    }

    if(value < 0.0f)
    {
        destTemp = -SPIRV_OCL_BUILTIN(fabs, _f32, )(destTemp);
    }

    return destTemp;
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ceil, _f32, )(float x ){
    return __builtin_IB_frnd_pi(x);
}

INLINE
float FDIV_IEEE( float a,
                           float b )
{
    return __builtin_IB_ieee_divide(a, b);
}

INLINE
double FDIV_IEEE_DOUBLE( double a,
                           double b )
{
    return __builtin_IB_ieee_divide_f64(a, b);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fabs, _f32, )(float x ){
    float neg = -x;
    return (x >= 0) ?  x : neg;
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(floor, _f32, )(float x ){
    return __builtin_IB_frnd_ni(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(rint, _f32, )(float x ){
    return __builtin_IB_frnd_ne(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(trunc, _f32, )(float x ){
    return __builtin_IB_frnd_zi(x);
}

INLINE
float FSQRT_IEEE( float a )
{
    return __builtin_IB_ieee_sqrt(a);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_cos, _f32, )(float x ){
    return __builtin_IB_native_cosf(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_exp2, _f32, )(float x ){
    return __builtin_IB_native_exp2f(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_log2, _f32, )(float x ){
    return __builtin_IB_native_log2f(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_powr, _f32_f32, )( float x,
                                           float y )
{
    return __builtin_IB_native_powrf(x, y);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_recip, _f32, )(float x ){
    return 1/x;
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_rsqrt, _f32, )(float x ){
    return SPIRV_OCL_BUILTIN(native_recip, _f32, )(SPIRV_OCL_BUILTIN(native_sqrt, _f32, )(x));
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_sin, _f32, )(float x ){
    return __builtin_IB_native_sinf(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_sqrt, _f32, )(float x ){
    return __builtin_IB_native_sqrtf(x);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_tan, _f32, )(float x ){
    return SPIRV_OCL_BUILTIN(native_divide, _f32_f32, )(SPIRV_OCL_BUILTIN(native_sin, _f32, )(x), SPIRV_OCL_BUILTIN(native_cos, _f32, )(x));
}

#ifdef cl_khr_fp16
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fclamp, _f16_f16_f16, )(half x, half minval, half maxval ){
    return SPIRV_OCL_BUILTIN(fmin, _f32_f32, )(SPIRV_OCL_BUILTIN(fmax, _f32_f32, )(x, minval), maxval);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(acos, _f16, )(half x ){
    return (half)SPIRV_OCL_BUILTIN(acos, _f32, )((float)x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(asin, _f16, )(half x ){
    return (half)SPIRV_OCL_BUILTIN(asin, _f32, )((float)x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(atan, _f16, )(half x ){
    return (half)SPIRV_OCL_BUILTIN(atan, _f32, )((float)x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ceil, _f16, )(half x ){
    return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(SPIRV_OCL_BUILTIN(ceil, _f32, )(SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x)));
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fabs, _f16, )(half x ){
    //return __builtin_IB_fabsh(x);
    return (x >= 0) ?  x : -x;

}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(floor, _f16, )(half x ){
    return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(SPIRV_OCL_BUILTIN(floor, _f32, )(SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x)));
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fma, _f16_f16_f16, )( half a,
                                      half b,
                                      half c )
{
    return __builtin_IB_fmah(a, b, c);
}

INLINE
float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fma, _f32_f32_f32, )( float a,
                                      float b,
                                      float c )
{
    return __builtin_fmaf(a, b, c);
}

INLINE
double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fma, _f64_f64_f64, )( double a,
                                      double b,
                                      double c )
{
    return __builtin_fma(a, b, c);
}


INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(mad, _f16_f16_f16, )( half a,
                                      half b,
                                      half c )
{
    //return __builtin_IB_madh(a, b, c);
    return a*b+c;
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(rint, _f16, )(half x ){
    return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(SPIRV_OCL_BUILTIN(rint, _f32, )(SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x)));
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(trunc, _f16, )(half x ){
    return SPIRV_BUILTIN(FConvert, _f16_f32, _Rhalf)(SPIRV_OCL_BUILTIN(trunc, _f32, )(SPIRV_BUILTIN(FConvert, _f32_f16, _Rfloat)(x)));
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_cos, _f16, )(half x ){
    return __builtin_IB_native_cosh(x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_exp2, _f16, )(half x ){
    return __builtin_IB_native_exp2h(x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_log2, _f16, )(half x ){
    return __builtin_IB_native_log2h(x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_powr, _f16_f16, )( half x,
                                          half y )
{
    return (half)__builtin_IB_native_powrf((float)x, (float)y);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_recip, _f16, )(half x ){
    return 1/x;
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_rsqrt, _f16, )(half x ){
    return SPIRV_OCL_BUILTIN(native_recip, _f16, )(SPIRV_OCL_BUILTIN(native_sqrt, _f16, )(x));
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_sin, _f16, )(half x ){
    return __builtin_IB_native_sinh(x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_sqrt, _f16, )(half x ){
    return __builtin_IB_native_sqrth(x);
}

INLINE
half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_tan, _f16, )(half x ){
    return SPIRV_OCL_BUILTIN(native_divide, _f32_f32, )(SPIRV_OCL_BUILTIN(native_sin, _f32, )((float)x), SPIRV_OCL_BUILTIN(native_cos, _f32, )((float)x));
}

#endif

#if defined(cl_khr_fp64)

INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fclamp, _f64_f64_f64, )(double x, double minval, double maxval ){
    return SPIRV_OCL_BUILTIN(fmin, _f64_f64, )(SPIRV_OCL_BUILTIN(fmax, _f64_f64, )(x, minval), maxval);
}

INLINE
double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(ceil, _f64, )(double x ){
    //First part of the algorithm performs rounding towards zero by truncating bits in
    //the fractional part of the number.This is done by finding out the position of the
    //fractional bits of the mantissa and masking them out with zeros.

    double roundedToZeroVal = SPIRV_OCL_BUILTIN(trunc, _f64, )(x);
    unsigned high32Bit = (int)(as_ulong( x ) >> 32);
    ulong fraction = as_ulong(x) - as_ulong(roundedToZeroVal);    // getting fraction

    // Second part is to calculate exponent adjustment based on sign of source

    uint temp5 = (uint)(as_ulong( fraction ));
    uint temp6 = (uint)(as_ulong( fraction ) >> 32);
    uint sign = high32Bit & 0x80000000;
    uint expBias = (sign == 0) ? 0x3ff00000 : 0;
    uint orDst = temp5 | temp6;
    ulong signAdjustedVal = (orDst == 0) ? 0 : (expBias);
    double output = as_double( signAdjustedVal << 32 ) + as_double( roundedToZeroVal );
    return output;
}

INLINE
double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(fabs, _f64, )(double x ){
    double neg = -x;
    return (x >= 0) ?  x : neg;
}

INLINE
double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(floor, _f64, )(double x ){
    //First part of the algorithm performs rounding towards zero by truncating bits in
    //the fractional part of the number.This is done by finding out the position of the
    //fractional bits of the mantissa and masking them out with zeros.

    double roundedToZeroVal = SPIRV_OCL_BUILTIN(trunc, _f64, )(x);

    ulong fraction = as_ulong(x) - as_ulong( roundedToZeroVal );       // getting fraction

    // Second part is to calculate exponent adjustment based on sign of source

    uint temp5 = (uint)( fraction );
    uint temp6 = (uint)(fraction >> 32);
    unsigned high32Bit = (uint)(as_ulong( x ) >> 32);
    uint sign = high32Bit & 0x80000000;
    uint expBias = (sign != 0) ? 0xbff00000 : 0;
    uint orDst = temp5 | temp6;
    ulong signAdjustedVal = (orDst == 0) ? 0 : (expBias);
    double output = as_double( signAdjustedVal << 32 ) + as_double( roundedToZeroVal );
    return output;
}

INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_sqrt, _f64, )(double x ){
        return __builtin_IB_native_sqrtd(x);
}

INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(native_rsqrt, _f64, )(double x ){
      return (1 / __builtin_IB_native_sqrtd(x));
}

#endif // defined(cl_khr_fp64)