VcExtraMath Class Reference

#include <vc_extra_math.h>

Static Public Member Functions
template<typename A >
static xsimd::batch< float, A >	erf (const xsimd::batch< float, A > x)

Detailed Description

Definition at line 15 of file vc_extra_math.h.

Member Function Documentation

erf()

template<typename A >

static xsimd::batch< float, A > VcExtraMath::erf ( const xsimd::batch< float, A > x )

inlinestatic

Our version of erf() is about 10% faster than the version in xsimd::erf(), because we require less precision. We don't need too much of precision to calculate 8-bit masks anyway.

Definition at line 21 of file vc_extra_math.h.

    {
        using float_v = xsimd::batch<float, A>;
        using float_m = typename float_v::batch_bool_type;
        float_v xa = xsimd::abs(x);
        float_m precisionLimit = xa >= float_v(9.3f); // wrong result for any number beyond this
        xa = xsimd::set_zero(xa, precisionLimit);
        float_v sign(1.0f);
        float_m invertMask = x < float_v(0.f);
        sign = xsimd::select(invertMask, float_v(-1.f), sign);
 
        // CONSTANTS
        float a1 =  0.254829592f;
        float a2 = -0.284496736f;
        float a3 =  1.421413741f;
        float a4 = -1.453152027f;
        float a5 =  1.061405429f;
        float p  =  0.3275911f;
 
        float_v t = 1.0f / (1.0f + p * xa);
        float_v y = 1.0f - (((((a5 * t + a4) * t) + a3) * t + a2) * t + a1) * t * exp(-xa * xa);
        y = xsimd::set_one(y, precisionLimit);
        return sign * y;
    }

References p, xsimd::set_one(), xsimd::set_zero(), and sign().

---

The documentation for this class was generated from the following file:

• libs/image/vc_extra_math.h