* The dec output exponent jumps in 8, so the result is offset at most * by 7 when the input is within range. */

* Ported from google double conversion strtod using * MathGeoLibs diy_fp functions for grisu3 in C. * * ulp_half_error is set if needed to trunacte

* `end` is unchanged if number is handled natively, or it is the result * of strtod parsing in case of fallback. */

Avoid dependence on lib math to get (int)ceil(v) */

* Returns end pointer on success, or null, or buf if start is not a number. * Sets result to 0.0 on error. * Reads up to len + 1 bytes from buffer w

* Returns buf if number wasn't matched, or null if number starts ok * but contains invalid content. */

From: https://stackoverflow.com/questions/1659440/32-bit-to-16-bit-floating-point-conversion