bend-collections

Search uses release 1.0.0.0 · 4,483 definitions

Releases (1)

1.0.0.0 indexed

Definitions (4,483)

src/math/f64.bend

def ld_cls
def ldexpx * 2^e, rounded once (overflow to infinity, gradual underflow)
def ulp_cls
def ulpthe value of the least significant bit of x (Python's math.ulp)
def with_sign
def na_step
def na_z
def na_eq
def na_nan
def nextafterthe next double after x toward y (IEEE nextAfter; y when x == y)
def fpick
def fmin_z
def fmin_ny
def fmin_nx
def fminIEEE 754-2019 minimumNumber: a NaN operand is ignored, -0 < +0
def fmax_z
def fmax_ny
def fmax_nx
def fmaxIEEE 754-2019 maximumNumber
def is_normal
def is_subnormal
def to_bits
def of_bits64
def ii_k
def ii_fin
def is_integerfinite with no fractional bits
def mf_frac
def mf_fin
def mf_cls
def modf(fractional part, integral part), both with the sign of x (Python's modf)
def fm_go(q, r) of mx * 2^d by b from (q0, r0) of mx by b: ten bits a step, so each partial remainder r * 2^t stays below 2^63
def fm_qrthe last quotient digit and remainder of mant(x) * 2^(e(x) - e(y)) by mant(y)
def fmod_fin
def fmod_z
def fmod_yi
def fmod_bad
def fmodx - n y for n = trunc(x / y), exact, with the sign of x (C's fmod)
def rm_pick
def rm_fixremainder r of divisor b at scale u, last quotient digit q: round the quotient to nearest (ties to even) by flipping to r - b
def rm_qr
def rm_near
def rm_far
def rm_z
def rm_yi
def rm_bad
def remainderIEEE remainder: x - n y for n = x / y rounded to nearest, ties to even
def ic_d
def ic_inf
def ic_eq
def ic_tol
def isclosePython's math.isclose (CPython's algorithm): a negative tolerance is a domain error
type Ratio+-num / 2^den in lowest terms (num odd or den = 0)
def ctz_gotrailing zeros of w (at most fuel), one halving at a time
def ctz
def ar_small
def ar_big
def ar_fin
def ar_z
def ar_top
def ar_cls
def as_integer_ratiox = +-num / 2^den exactly, in lowest terms (Python's as_integer_ratio with the denominator's exponent); NaN is a domain error, infinities and integers of 2^64 or more overflow
def f64_op
def f64_is

src/math/generic.bend

type QuotRem
def ok
def or_zero
def fuelenough steps for every loop at 64 bits (Euclid < 1.5 w + 2, bit loops w)
def fitsx when it fits, else None
def cadd
def cmul
def lt
def le
def pos
def pick
def minPython's min(a, b) keeps a unless b < a; max keeps a unless a < b.
def max
def clamp_ok
def clampmin(max(x, lo), hi); hi < lo is a Domain error
def abs
def sign_below
def sign_above
def sign1 above zero, -1 below, x itself otherwise (0, -0.0, NaN)
def sum_go
def sum
def prod_go
def prod
def mul_st(value, still fits) after multiplying by x when bit is set
def sq_valthe base squared, only while higher exponent bits remain
def sq_ok
def st_fin
def pow_gosquare-and-multiply, low bits first; an overflowing square means the result overflows too (a later bit multiplies it in).
def pow
def one
def inc
def dec
def zst(x, x == 0)
def gcd_goEuclid's algorithm on (a, (b, b == 0))
def hst
def strip_gox with its factors of two removed (x != 0)
def strip