Lines Matching refs:precision
37 * The precision of the mathematical operations on 32 bit floats is affected by the pragmas
43 * Different precision/speed tradeoffs can be achieved by using variants of the common math
45 * - native_: May have custom hardware implementations with weaker precision. Additionally,
1706 * This function is similar to mad(). fma() retains full precision of the multiplied result
1708 * This extra precision is not guaranteed in rs_fp_relaxed mode.
2071 * half_recip: Reciprocal computed to 16 bit precision
2075 * The precision is that of a 16 bit floating point value.
2100 * half_rsqrt: Reciprocal of a square root computed to 16 bit precision
2104 * The precision is that of a 16 bit floating point value.
2129 * half_sqrt: Square root computed to 16 bit precision
2133 * The precision is that of a 16 bit floating point value.
2588 * This function is similar to fma(). fma() retains full precision of the multiplied result
3670 * Returns a half-precision floating point NaN value (Not a Number).
4459 * It is valid for inputs from -86.f to 86.f. The precision is no worse than what would be
4509 * It is valid for inputs from -37.f to 37.f. The precision is no worse than what would be
4559 * It is valid for inputs from -125.f to 125.f. The precision is no worse than what would be