core::arch::x86

Function _mm512_mask_getmant_round_ph

Source
pub unsafe fn _mm512_mask_getmant_round_ph(
    src: __m512h,
    k: __mmask32,
    a: __m512h,
    const NORM: _MM_MANTISSA_NORM_ENUM,
    const SIGN: _MM_MANTISSA_SIGN_ENUM,
    const SAE: i32,
) -> __m512h
🔬This is a nightly-only experimental API. (stdarch_x86_avx512_f16 #127213)
Available on (x86 or x86-64) and target feature avx512fp16 and x86 only.
Expand description

Normalize the mantissas of packed half-precision (16-bit) floating-point elements in a, and store the results in dst using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates ±(2^k)*|x.significand|, where k depends on the interval range defined by norm and the sign depends on sign and the source sign. Exceptions can be suppressed by passing _MM_FROUND_NO_EXC in the sae parameter

The mantissa is normalized to the interval specified by interv, which can take the following values:

_MM_MANT_NORM_1_2     // interval [1, 2)
_MM_MANT_NORM_p5_2    // interval [0.5, 2)
_MM_MANT_NORM_p5_1    // interval [0.5, 1)
_MM_MANT_NORM_p75_1p5 // interval [0.75, 1.5)

The sign is determined by sc which can take the following values:

_MM_MANT_SIGN_src     // sign = sign(src)
_MM_MANT_SIGN_zero    // sign = 0
_MM_MANT_SIGN_nan     // dst = NaN if sign(src) = 1

Exceptions can be suppressed by passing _MM_FROUND_NO_EXC in the sae parameter

Intel’s documentation