What is the fastest method to shift matrix?

[Amro]

Thank you very much for your tips.

@acxz

I’d just test the permute solution for your workload @Amro and see if the timings are better on your specific hardware/workload.

@acxz

# Permute
    A[:, 1] .= 0
    dropzeros!(A)
    permute!(A, 1:size(A, 1), vcat(2:size(A, 2), 1))

I have tried the permute! with dropzeros! but it is slower. Yes, I think I should not apply the dropzeros! at each time, but this is a bit annoying.