Everyone loves Krylov(.jl)
Running fft_versus_pcg_solver.jl produces:
julia> include("fft_versus_pcg_solver.jl")
0.002864 seconds (26 allocations: 6.688 KiB)
21.955589 seconds (410.61 M allocations: 13.020 GiB, 3.33% gc time)
[ Info: PCG solver iterations: 190
[ Info: Max PCG residual: 1.277806338387602e-9and the figure
The PCG solver seems to perform very poorly and allocates way too much.
I'm not sure why this might be, but it could be an issue with norm.