Float64x4 (ocaml_simd.ocaml_simd.avx.Ocaml_simd

type t = float64x4

type mask = int64x4

val box : t -> float64x4 @@ portable

val unbox : float64x4 @ local -> t @@ portable

val zero : unit -> t @@ portable

Equivalent to const1 #0.0.

val one : unit -> t @@ portable

Equivalent to const1 #1.0.

val set1 : float -> t @@ portable

_mm256_set1_pd

val set : float -> float -> float -> float -> t @@ portable

_mm256_set_pd

val set_lanes : float64x2 -> float64x2 -> t @@ portable

_mm256_set_m128 Operates on two float64x2 lanes.

val const1 : float -> t @@ portable

Argument must be a literal or an unboxing function applied to a literal. Compiles to a static vector literal. Exposed as an external so user code can compile without cross-library inlining.

val const : float -> float -> float -> float -> t @@ portable

Arguments must be literals or unboxing functions applied to literals. Compiles to a static vector literal. Exposed as an external so user code can compile without cross-library inlining.

module Raw : sig ... end

module String : sig ... end

module Bytes : sig ... end

module Bigstring : sig ... end

module Float_array : sig ... end

module Float_iarray : sig ... end

module Floatarray : sig ... end

module Float_u_array : sig ... end

val (>=) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_LE_OS

val (<=) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_LE_OS

val (=) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_EQ_OQ

val (>) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_LT_OS

val (<) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_LT_OS

val (<>) : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_NEQ_UQ

val equal : t -> t -> mask @@ portable

_mm256_cmp_pd with _CMP_EQ_OQ

val is_nan : t -> mask @@ portable

_mm256_cmp_pd with _CMP_UNORD_Q

val is_not_nan : t -> mask @@ portable

_mm256_cmp_pd with _CMP_ORD_Q

val movemask : mask -> int64 @@ portable

_mm256_movemask_pd

val bitmask : mask -> int64x4 @@ portable

Identity.

val select : mask -> fail:t -> pass:t -> t @@ portable

_mm256_blendv_pd Only reads the sign bit of each mask lane. Selects the element from pass if the sign bit is 1, otherwise fail.

val insert : idx:int64 -> t -> float -> t @@ portable

idx must be in 0,3.

val extract : idx:int64 -> t -> float @@ portable

idx must be in 0,3.

val extract0 : t -> float @@ portable

Projection. Has no runtime cost.

val insert_lane : idx:int64 -> t -> float64x2 -> t @@ portable

idx must be a literal in 0,1. Operates on two float64x2 lanes. Exposed as an external so user code can compile without cross-library inlining.

val extract_lane : idx:int64 -> t -> float64x2 @@ portable

idx must be a literal in 0,1. Operates on two float64x2 lanes. Exposed as an external so user code can compile without cross-library inlining.

val extract_lane0 : t -> float64x2 @@ portable

Projection. Has no runtime cost. Operates on two float64x2 lanes.

val splat : t -> #(float * float * float * float) @@ portable

Slow, intended for debugging / printing / etc.

val interleave_upper_lanes : even:t -> odd:t -> t @@ portable

_mm256_unpackhi_pd Operates on two float64x2 lanes.

  interleave_upper_lanes ~even ~odd = (even.(1), odd.(1), even.(3), odd.(3))

val interleave_lower_lanes : even:t -> odd:t -> t @@ portable

_mm256_unpacklo_pd Operates on two float64x2 lanes.

  interleave_lower_lanes ~even ~odd = (even.(0), odd.(0), even.(2), odd.(2))

val duplicate_even : t -> t @@ portable

_mm256_movedup_pd

  duplicate_even x = (x.(0), x.(0), x.(2), x.(2))

val blend : Ocaml_simd.Blend4.t -> t -> t -> t @@ portable

_mm256_blend_pd Specify blend with ppx_simd: %blend N, N, N, N, where each N is in 0,1. Exposed as an external so user code can compile without cross-library inlining.

  blend [%blend 1, 0, 1, 0] x y = (y.(0), x.(1), y.(2), x.(3))

val shuffle_lanes : Ocaml_simd.Shuffle2x2.t -> t -> t -> t @@ portable

_mm256_shuffle_pd Specify shuffle with ppx_simd: %shuffle (N, N), (N, N), where each N is in 0,1. Operates on two float64x2 lanes. Exposed as an external so user code can compile without cross-library inlining.

  shuffle_lanes [%shuffle (1, 0), (0, 1)] x y = (x.(1), y.(0), x.(2), y.(3))

val permute : Ocaml_simd.Permute4.t -> t -> t @@ portable

_mm256_permute4x64_pd Specify permute with ppx_simd: %permute N, N, N, N, where each N is in 0,3. Exposed as an external so user code can compile without cross-library inlining.

  permute [%permute 3, 1, 0, 2] x = (x.(3), x.(1), x.(0), x.(2))

val permute_lanes : Ocaml_simd.Permute2x2.t -> t -> t @@ portable

_mm256_permute_pd Specify permute with ppx_simd: %permute (N, N), (N, N), where each N is in 0,1. Exposed as an external so user code can compile without cross-library inlining.

  permute_lanes [%permute (1, 0), (0, 1)] x = (x.(1), x.(0), x.(2), x.(3))

val permute_lanes_by : t -> idx:int64x4 -> t @@ portable

_mm256_permutevar_pd Operates on two float64x2 lanes. Each lane of idx is interpreted as an integer in 0,1 by taking its *second* lowest bit.

  permute_lanes_by x ~idx:(2, 0, 0, 2) = (x.(1), x.(0), x.(2), x.(3))

val min : t -> t -> t @@ portable

_mm256_min_pd Equivalent to pointwise (x < y ? x : y). If either lane is NaN, the second lane is returned.

val max : t -> t -> t @@ portable

_mm256_max_pd Equivalent to pointwise (x > y ? x : y). If either lane is NaN, the second lane is returned.

val add : t -> t -> t @@ portable

_mm256_add_pd

val sub : t -> t -> t @@ portable

_mm256_sub_pd

val mul : t -> t -> t @@ portable

_mm256_mul_pd

val div : t -> t -> t @@ portable

_mm256_div_pd

val neg : t -> t @@ portable

Compiles to xor with a static constant.

val abs : t -> t @@ portable

Compiles to and with a static constant.

val sqrt : t -> t @@ portable

_mm256_sqrt_pd

val add_sub : t -> t -> t @@ portable

_mm256_addsub_pd

  add_sub x y = (x.(0) - y.(0), x.(1) + y.(1), x.(2) - y.(2), x.(3) + y.(3))

val horizontal_add_lanes : t -> t -> t @@ portable

_mm256_hadd_pd Operates on two float64x2 lanes.

  horizontal_add_lanes x y
  = (x.(1) + x.(0), x.(3) + x.(2), y.(1) + y.(0), y.(3) + y.(2))

val horizontal_sub_lanes : t -> t -> t @@ portable

_mm256_hsub_pd Operates on two float64x2 lanes.

  horizontal_sub_lanes x y
  = (x.(0) - x.(1), x.(2) - x.(3), y.(0) - y.(1), y.(2) - y.(3))

val mul_add : t -> t -> t -> t @@ portable

_mm256_fmadd_pd. Computes x * y + z without intermediate rounding.

val mul_sub : t -> t -> t -> t @@ portable

_mm256_fmsub_pd. Computes x * y - z without intermediate rounding.

val neg_mul_add : t -> t -> t -> t @@ portable

_mm256_fnmadd_pd. Computes -(x * y) + z without intermediate rounding.

val neg_mul_sub : t -> t -> t -> t @@ portable

_mm256_fnmsub_pd. Computes -(x * y) - z without intermediate rounding.

val mul_add_sub : t -> t -> t -> t @@ portable

_mm256_fmaddsub_ps. Computes the following expression without intermediate rounding.

  mul_add_sub x y z
  = ( (x.(0) * y.(0)) - z.(0)
    , (x.(1) * y.(1)) + z.(1)
    , (x.(2) * y.(2)) - z.(2)
    , (x.(3) * y.(3)) + z.(3) )

val mul_sub_add : t -> t -> t -> t @@ portable

_mm256_fmsubadd_ps. Computes the following expression without intermediate rounding.

  mul_add_sub x y z
  = ( (x.(0) * y.(0)) + z.(0)
    , (x.(1) * y.(1)) - z.(1)
    , (x.(2) * y.(2)) + z.(2)
    , (x.(3) * y.(3)) - z.(3) )

val (+) : t -> t -> t @@ portable

val (-) : t -> t -> t @@ portable

val (/) : t -> t -> t @@ portable

val (*) : t -> t -> t @@ portable

val iround_current : t -> int32x4 @@ portable

_mm256_cvtpd_epi32

val round_nearest : t -> t @@ portable

_mm256_round_pd

val round_current : t -> t @@ portable

_mm256_round_pd

val round_down : t -> t @@ portable

_mm256_round_pd

val round_up : t -> t @@ portable

_mm256_round_pd

val round_toward_zero : t -> t @@ portable

_mm256_round_pd

val unsafe_of_float : float -> t @@ portable

Identity; leaves upper 192 bits unspecified.

val unsafe_of_float64x2 : float64x2 -> t @@ portable

Identity; leaves upper 128 bits unspecified.

val of_float16x16_bits : float16x16 -> t @@ portable