1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2017-2021 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
#include "../include/BiF_Definitions.cl"
#include "../../Headers/spirv.h"
INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(smoothstep, _f32_f32_f32, )(float edge0, float edge1, float x ){
float div = (x - edge0) / (edge1 - edge0);
float temp = SPIRV_OCL_BUILTIN(fclamp, _f32_f32_f32, )( div, 0.0f, 1.0f );
return temp * temp * SPIRV_OCL_BUILTIN(mad, _f32_f32_f32, )( -2.0f, temp, 3.0f );
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_3ARGS( smoothstep, float, float, f32 )
#if defined(cl_khr_fp64)
INLINE double SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(smoothstep, _f64_f64_f64, )(double edge0, double edge1, double x ){
double div = (x - edge0) / (edge1 - edge0);
double temp = SPIRV_OCL_BUILTIN(fclamp, _f64_f64_f64, )( div, 0.0, 1.0 );
return temp * temp * SPIRV_OCL_BUILTIN(mad, _f64_f64_f64, )( -2.0, temp, 3.0 );
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_3ARGS( smoothstep, double, double, f64 )
#endif // defined(cl_khr_fp64)
#if defined(cl_khr_fp16)
INLINE half SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(smoothstep, _f16_f16_f16, )(half edge0, half edge1, half x ){
half div = (x - edge0) / (edge1 - edge0);
half temp = SPIRV_OCL_BUILTIN(fclamp, _f16_f16_f16, )( div, (half)0.0f, (half)1.0f );
return temp * temp * SPIRV_OCL_BUILTIN(mad, _f16_f16_f16, )( (half)-2.0f, temp, (half)3.0f );
}
GENERATE_SPIRV_OCL_VECTOR_FUNCTIONS_3ARGS( smoothstep, half, half, f16 )
#endif // defined(cl_khr_fp16)
|