hmm.jl

using BenchmarkTools
A = rand(10000)
B = Vector[rand(10) for i = 1:1000]
C = rand(10,1000)

D = rand(10000)
E = Vector[rand(10) for i = 1:1000]
F = rand(10, 1000)

function fast!(a, c)
    for j = 1:1000
        a[1+(j-1)*10:j*10] = c[1+(j-1)*10:j*10]'*rand(10,10)
    end
end

function fast_slow!(c, f)
    for j = 1:1000
        c[:, j] = f[:, j]'*rand(10,10)
    end
end

function slow!(b, d)
    for j = 1:1000
            b[j][:] = d[j]'*rand(10,10)
    end
end

fast!(A, D)
fast_slow!(C, F)
slow!(B, E)
@benchmark fast!(A, C)
@benchmark fast_slow!(C, F)
@benchmark slow!(B, E)

BenchmarkTools.Trial: 
  samples:          6536
  evals/sample:     1
  time tolerance:   5.00%
  memory tolerance: 1.00%
  memory estimate:  1.34 mb
  allocs estimate:  5000
  minimum time:     681.45 μs (0.00% GC)
  median time:      700.28 μs (0.00% GC)
  mean time:        763.74 μs (7.17% GC)
  maximum time:     2.50 ms (61.59% GC)

julia> @benchmark fast_slow!(C, F)
BenchmarkTools.Trial: 
  samples:          5794
  evals/sample:     1
  time tolerance:   5.00%
  memory tolerance: 1.00%
  memory estimate:  1.44 mb
  allocs estimate:  9467
  minimum time:     743.66 μs (0.00% GC)
  median time:      755.04 μs (0.00% GC)
  mean time:        861.39 μs (11.62% GC)
  maximum time:     2.71 ms (68.97% GC)

julia> @benchmark slow!(B, E)
BenchmarkTools.Trial: 
  samples:          6381
  evals/sample:     1
  time tolerance:   5.00%
  memory tolerance: 1.00%
  memory estimate:  1.19 mb
  allocs estimate:  4000
  minimum time:     718.93 μs (0.00% GC)
  median time:      730.01 μs (0.00% GC)
  mean time:        782.54 μs (6.26% GC)
  maximum time:     2.17 ms (60.21% GC)