time_SVector_mul.jl

using LinearAlgebra
using StaticArrays
using BenchmarkTools

function time_SVector_3arg_mul(b, A, x)
    mul!(b, A, x)
end
function time_SVector_5arg_mul(b, A, x, alpha, beta)
    mul!(b, A, x, alpha, beta)
end

function time_Vector_3arg_mul(b1, b2, A, x1, x2)
    mul!(b1, A, x1)
    mul!(b2, A, x2)
end

function time_Vector_5arg_mul(b1, b2, A, x1, x2, alpha, beta)
    mul!(b1, A, x1, alpha, beta)
    mul!(b2, A, x2, alpha, beta)
end

for N in [10 15 20 35]
    A = randn(N, N)
    x = [SVector(1.0, 2.0) for _ in axes(A, 2)]
    b = similar(x)
    x1, x2 = [getindex.(x, i) for i in 1:2]
    b1, b2 = similar.((x1, x2))

    println("==== On N = $N ====")
    println("SVector mul! timings are")
    @btime time_SVector_3arg_mul($b, $A, $x)
    @btime time_SVector_5arg_mul($b, $A, $x, $.1, $.5)

    println("Vector mul! timings are")
    @btime time_Vector_3arg_mul($b1, $b2, $A, $x1, $x2)
    @btime time_Vector_5arg_mul($b1, $b2, $A, $x1, $x2, $.1, $.5)

    println("")
end

# Results:
# ==== On N = 10 ====
# SVector mul! timings are
#   57.977 ns (0 allocations: 0 bytes)
#   73.356 ns (1 allocation: 32 bytes)
# Vector mul! timings are
#   140.434 ns (0 allocations: 0 bytes)
#   137.862 ns (0 allocations: 0 bytes)

# ==== On N = 15 ====
# SVector mul! timings are
#   123.608 ns (0 allocations: 0 bytes)
#   140.637 ns (1 allocation: 32 bytes)
# Vector mul! timings are
#   214.298 ns (0 allocations: 0 bytes)
#   212.368 ns (0 allocations: 0 bytes)

# ==== On N = 20 ====
# SVector mul! timings are
#   202.477 ns (0 allocations: 0 bytes)
#   219.212 ns (1 allocation: 32 bytes)
# Vector mul! timings are
#   215.278 ns (0 allocations: 0 bytes)
#   209.742 ns (0 allocations: 0 bytes)

# ==== On N = 35 ====
# SVector mul! timings are
#   578.066 ns (0 allocations: 0 bytes)
#   594.737 ns (1 allocation: 32 bytes)
# Vector mul! timings are
#   485.608 ns (0 allocations: 0 bytes)
#   496.778 ns (0 allocations: 0 bytes)