Skip to content

fft benchmark problems #2

Description

@stevengj

You wrote:
image
However, for such a small transform, you're seeing a lot of overhead in fft(a) from setting up the FFT plan, setting up the trig tables, etcetera. To really get the full benefit of FFTW, you need to create a precomputed plan:

julia> p = plan_fft(a);

julia> @time b3 = p*a;
  0.002879 seconds (1.39 k allocations: 97.771 KB)

julia> @time b3 = p*a;
  0.000026 seconds (10 allocations: 32.375 KB)

On my machine, my_fft(a) is about 6 times lower than fft(a), but about 33 times slower than p*a.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions