Function cpf_powf

Function Documentation

static inline int cpf_powf(float *X, const float *A, const float *B, const size_t numelem, optstruct *fpopts)

Compute real powers in lower precision.

If the function executes without errors, then \( X_i = A_i^{B_i} \) rounded to a lower-precision target format. The parameters of the target format and the rounding mode to be used are encoded in fpopts. If required, the function flips one bit in some of the entries of X.

If OpenMP support is specified at compile time, several OpenMP threads are used if numelem is large enough. This parameter is machine dependent.

Parameters:

  • X – [out] Array of rounded values.

  • A – [in] Input array.

  • B – [in] Input array.

  • numelem – [in] Number of elements in X, A, and B.

  • fpopts – [in] Parameters describing the target format, the rounding mode, and the probability of soft errors striking the rounded values.

Returns:

The function returns 1 if fpopts->precision is larger than 24 , 2 if fptops->emin is smaller than -126 or fpopts->emax is larger than 127 , and 0 otherwise.