Clarify rounding on VEC_UNSIGNED*, VEC_SIGNED*, and VEC_FLOAT* description #21

VEC_UNSIGNED*, VEC_SIGNED* description isn't clear on how the FP values are converted to integer values.
Similarly, for VEC_FLOATE, and FLOATO, it isn't clear on the rounding used to convert from vector double to vector float.

From what I obtained from Ian:

vec_signed and vec_unsigned round toward zero. If the fp value can't be represented because it's too low or too high then they saturate to the smallest or largest valid integer.

vec_floate and vec_floato round to the current rounding mode RN in the FPSCR.

VEC_UNSIGNED*, VEC_SIGNED* description isn't clear on how the FP values are converted to integer values. Similarly, for VEC_FLOATE, and FLOATO, it isn't clear on the rounding used to convert from vector double to vector float. From what I obtained from Ian: vec_signed and vec_unsigned round toward zero. If the fp value can't be represented because it's too low or too high then they saturate to the smallest or largest valid integer. vec_floate and vec_floato round to the current rounding mode RN in the FPSCR.

Looks like we clarified this for vec_signed and vec_unsigned, but not for the others.

Labels Milestones

Clarify rounding on VEC_UNSIGNED, VEC_SIGNED, and VEC_FLOAT* description #21