benchmark.jl

using BenchmarkTools, Plots, TunedSmoother
using GeometricMultigrid: Vcycle!

data = create_synthetic(;n=64)
begin
    smoothers = [Jacobi!,GS!,SOR!,pseudo!]
    suite = BenchmarkGroup()
    for s ∈ smoothers, (name,d) ∈ data
        suite[name,s] = @benchmarkable Vcycle!(st;smooth! = $s) setup=(st=state($d[1]...;smooth! = $s))
    end
end
begin # run benchmarks and count iterations
    results = run(suite)
    times = [minimum(results[name,s]).time for s ∈ smoothers, name ∈ keys(data)]
    time_factor = times[2:end,:]./times[1,:]'
end

counts = [avecount(d,s;p=TunedSmoother.p₀) for s ∈ smoothers[2:end], (_,d) ∈ data]

begin # results from block above
    time_factor = [ 3.50716   2.72302   3.5       2.73946   3.51437   2.71142
                    3.19072   3.12036   3.19      3.12826   3.1877    3.13066
                    1.37096   1.56152   1.36767   1.54179   1.36671   1.54011]
    counts = [  3.0  4.0  1.58  2.04  3.0  3.0
                4.0  3.0  1.7   2.05  3.0  3.9
                3.0  3.0  1.57  2.04  3.16  3.0]
    rel_time = counts.*time_factor
    accel = rel_time[1:2,:]./rel_time[3,:]' # speed up!!
end

begin # plot
    using StatsPlots,CategoricalArrays
    cats = ["GS","SOR","Ã⁻¹"]
    data = create_synthetic(len=1)
    n = length(data)
    colors = repeat([palette(:default)[2],palette(:default)[4],palette(:default)[3]],inner=n)
    ctg = CategoricalArray(repeat(cats,inner=n))
    levels!(ctg,cats)
    groupedbar(rel_time',size=(500,400),
                group=ctg,c=colors,legend=:outerbottomright, 
                yaxis=("relative time"),
                xaxis=("cases",(1:n,keys(data)),60))
end
savefig("synthetic_timing.pdf")