simdpp::rcp_e

 

libsimdpp

Types

Capabilities

Operations

Conversions

Bitwise operations

Floating-point operations

Integer operations

Memory access operations

Shuffle operations

Miscellaneous operations

 

Floating-point operations

Elementary operations
add sub	mul div	fmadd fmsub	neg
Rounding
round ceil	floor trunc
Comparisons
cmp_eq cmp_neq	cmp_lt cmp_le	cmp_gt cmp_ge
Reduction
reduce_add reduce_min	reduce_mul reduce_max
Miscellaneous
abs sign	min max	isnan isnan2
Square root and reciprocal
sqrt	rcp_e rcp_rh	rsqrt_e rsqrt_rh

 

template<unsigned N, class E> Ret<N, _DETAIL_> rcp_e( Vec<N,E> a );

Computes approximate reciprocal. The implementation behaves as if the following set of overloads is provided:

Ret	Vec
float32	float32
float64	float64

The precision depends on platform:

• 1/2 ULP for NULL and NEON
• ~1/2730 for SSE2
• 1/16376 for AVX512F
• 1/4096 for ALTIVEC
• 1/256 for NEON_FLT_SP

Parameters

a

-

floating-point vector

Return value

A vector expression evaluating to approx(1/a).

Equivalent operation

r0 = approx(1 / a0)
...
rN = approx(1 / aN)