Module `simd`

SIMD compiler intrinsics.

In this module, a "vector" is any repr(simd) type.

Enums

SimdAlign A type for alignment options for SIMD masked load/store intrinsics.

Functions

simd_add Adds two simd vectors elementwise.
simd_and "And"s vectors elementwise.
simd_arith_offset Calculates the offset from a pointer vector elementwise, potentially wrapping.
simd_as Numerically casts a vector, elementwise.
simd_bitmask Truncates an integer vector to a bitmask.
simd_bitreverse Reverses bits of each element.
simd_bswap Swaps bytes of each element.
simd_cast Numerically casts a vector, elementwise.
simd_cast_ptr Casts a vector of pointers.
simd_ceil Rounds up each element to the next highest integer-valued float.
simd_ctlz Counts the leading zeros of each element.
simd_ctpop Counts the number of ones in each element.
simd_cttz Counts the trailing zeros of each element.
simd_div Divides lhs by rhs elementwise.
simd_eq Tests elementwise equality of two vectors.
simd_expose_provenance Exposes a vector of pointers as a vector of addresses.
simd_extract Extracts an element from a vector.
simd_extract_dyn Extracts an element from a vector.
simd_fabs Returns absolute value of a vector, elementwise.
simd_fcos T must be a vector of floats.
simd_fexp T must be a vector of floats.
simd_fexp2 T must be a vector of floats.
simd_flog T must be a vector of floats.
simd_flog10 T must be a vector of floats.
simd_flog2 T must be a vector of floats.
simd_floor Rounds down each element to the next lowest integer-valued float.
simd_fma Computes (x*y) + z for each element, but without any intermediate rounding.
simd_fmax Returns the maximum of two vectors, elementwise.
simd_fmin Returns the minimum of two vectors, elementwise.
simd_fsin T must be a vector of floats.
simd_fsqrt Takes the square root of each element.
simd_funnel_shl Funnel Shifts vector left elementwise, with UB on overflow.
simd_funnel_shr Funnel Shifts vector right elementwise, with UB on overflow.
simd_gather Reads a vector of pointers.
simd_ge Tests if x is greater than or equal to y, elementwise.
simd_gt Tests if x is greater than y, elementwise.
simd_insert Inserts an element into a vector, returning the updated vector.
simd_insert_dyn Inserts an element into a vector, returning the updated vector.
simd_le Tests if x is less than or equal to y, elementwise.
simd_lt Tests if x is less than y, elementwise.
simd_masked_load Reads a vector of pointers.
simd_masked_store Writes to a vector of pointers.
simd_mul Multiplies two simd vectors elementwise.
simd_ne Tests elementwise inequality equality of two vectors.
simd_neg Negates a vector elementwise.
simd_or "Ors" vectors elementwise.
simd_reduce_add_ordered Adds elements within a vector from left to right.
simd_reduce_add_unordered Adds elements within a vector in arbitrary order. May also be re-associated with unordered additions on the inputs/outputs.
simd_reduce_all Checks if all mask values are true.
simd_reduce_and Logical "and"s all elements together.
simd_reduce_any Checks if any mask value is true.
simd_reduce_max Returns the maximum element of a vector.
simd_reduce_min Returns the minimum element of a vector.
simd_reduce_mul_ordered Multiplies elements within a vector from left to right.
simd_reduce_mul_unordered Multiplies elements within a vector in arbitrary order. May also be re-associated with unordered additions on the inputs/outputs.
simd_reduce_or Logical "ors" all elements together.
simd_reduce_xor Logical "exclusive ors" all elements together.
simd_relaxed_fma Computes (x*y) + z for each element, non-deterministically executing either a fused multiply-add or two operations with rounding of the intermediate result.
simd_rem Returns remainder of two vectors elementwise.
simd_round Rounds each element to the closest integer-valued float. Ties are resolved by rounding away from 0.
simd_round_ties_even Rounds each element to the closest integer-valued float. Ties are resolved by rounding to the number with an even least significant digit
simd_saturating_add Adds two simd vectors elementwise, with saturation.
simd_saturating_sub Subtracts two simd vectors elementwise, with saturation.
simd_scatter Writes to a vector of pointers.
simd_select Selects elements from a mask.
simd_select_bitmask Selects elements from a bitmask.
simd_shl Shifts vector left elementwise, with UB on overflow.
simd_shr Shifts vector right elementwise, with UB on overflow.
simd_shuffle Shuffles two vectors by const indices.
simd_sub Subtracts rhs from lhs elementwise.
simd_trunc Returns the integer part of each element as an integer-valued float. In other words, non-integer values are truncated towards zero.
simd_with_exposed_provenance Creates a vector of pointers from a vector of addresses.
simd_xor "Exclusive ors" vectors elementwise.