JeffreySarnoff

struct ZERO end
const Zero = ZERO()

# to display Zero as Zero
const ShowZero = :Zero
# to display Zero as 0
# const ShowZero = 0
Base.show(io::IO, x::ZERO) = print(io, ShowZero)

Base.zero(::ZERO) = 0

JeffreySarnoff

# very small subset of data, for testing
       
N = 5 # people
T = 3 # times

data = Float32.([
#    x     y      z
    1.75   0.0    0.75  # person 1 time 1
    2.75   1.75   0.75  # person 1 time 2
    1.25   2.75  -0.75  # person 1 time 3

JeffreySarnoff

# see example at line 80

using StatsBase

function StatsBase.median(a::T, b::T, c::T) where {T}
    b, c = minmax(b, c)
    a, c = minmax(a, c)
    max(a, b)
end

JeffreySarnoff

# see example at line 70
# see accumulator enhancement at line 70

using StatsBase

abstract type Accumulator{T} <: Function end

const AccNum = Float64 # client settable via ENV

StatsBase.nobs(acc::Accumulator) = acc.nobs

JeffreySarnoff

# see example at line 60

using StatsBase

abstract type Accumulator{T} <: Function end

const AccNum = Float64 # client settable via ENV

StatsBase.nobs(acc::Accumulator) = acc.nobs

JeffreySarnoff

function shiftup(v::AbstractVector{T}, start, finish) where {T}
    for i in finish:-1:start
        @inbounds v[i+1] = v[i]
    end
end

function shiftup4(v::AbstractVector{T}, start, finish) where {T}
    n = finish - start
    q, r = fldmod(n, 4)
    @inbounds for i in finish-3:-4:start+r

JeffreySarnoff

#=
 * Data for log.
 *
 * Copyright (c) 2018, Arm Limited.
 * SPDX-License-Identifier: MIT OR Apache-2.0 WITH LLVM-exception
 =#

const OneAsUInt64 = 0x3ff0000000000000

const LOG_TABLE_BITS = 7

JeffreySarnoff

using BenchmarkTools
const BT=BenchmarkTools.DEFAULT_PARAMETERS;
BT.samples=60_000; BT.time_tolerance=1/8192;
#=
IS{Int64} == Union{Int64, Symbol}
IS{Int32} == Union{Int32, Symbol}
=#
const IS = Union{I,Symbol} where {I<:Integer}

# test vars, length(chrs) == 96

JeffreySarnoff

re: https://twitter.com/SimonDeDeo/status/1344031423704006656

You are not timing the same thing in each language. See below for what you were timing in Julia.
And, as has been mentioned, use @time or @btime from BenchmarkTools.jl (time does not do what you expect it to do). When using @time, run the timing once to precompile the test before considering the results valid. The first run may take considerably longer, and that extra time is not computing time. I have omitted those lines below.

JeffreySarnoff

# provide context when a NaN is generated
# IMPORTANT!! define these before overloading the arithmetic functions

function nanfound(fncall)
    stk = stacktrace() # stacktrace(catch_backtrace())
    stk = relevantframes(stk)
    info = []
    for frame in stk
        push!(info, frameinfo(frame))
    end