* For q in (0,350), we have that * f = (((152170 + 65536) * q ) >> 16); * is equal to * floor(p) + q * where * p = log(5**q)/log(2) = q * log(5)/log(2) * * For negative values of q in (-400,0), we have that * f = (((152170 + 65536) * q ) >> 16); * is equal to * -ceil(p) + q * where * p = log(5**-q)/log(2) = -q * log(5)/log(2) */
| 1685 | * p = log(5**-q)/log(2) = -q * log(5)/log(2) |
| 1686 | */ |
| 1687 | constexpr fastfloat_really_inline int32_t power(int32_t q) noexcept { |
| 1688 | return (((152170 + 65536) * q) >> 16) + 63; |
| 1689 | } |
| 1690 | } // namespace detail |
| 1691 | |
| 1692 | // create an adjusted mantissa, biased by the invalid power2 |
no outgoing calls
no test coverage detected