Thread overview
fft and isPowerOf2?
May 17
kdevel
May 17
Hi,

this applications throws an error in std.numeric (Line 2826).
=> assert(isPowerOf2(range.length));

-----------
module std.numeric

void fftImplPureReal(Ret, R)(R range, Ret buf) const
    in
    {
        assert(range.length >= 4);
        assert(isPowerOf2(range.length));
    }
-----------

The application is rewritten from python (numpy).
How can I workaround this issue? Isn't it possible to give an arbitrary length
of data to fft like in numpy?

import std.numeric;
import std.datetime.stopwatch;
import std.stdio: writeln;
import std.random;
import std.array : array;
import std.range : generate, takeExactly;

void fourierTest(int samplingRate)
{
    auto sw = StopWatch(AutoStart.yes);

    foreach(n; 0..60 * 24)
    {
        sw.stop();
        double[] x = generate!(() => uniform01).takeExactly(samplingRate * 60).array;
        sw.start();

        writeln(x);
        auto y = fft(x);
    }

    sw.stop();
    writeln("microseconds: ", sw.peek.total!"usecs");
}

void main()
{
    fourierTest(50);
}

Kind regards
André
May 17
On Thursday, 17 May 2018 at 12:34:25 UTC, Andre Pany wrote:
> this applications throws an error in std.numeric (Line 2826).
> => assert(isPowerOf2(range.length));

> Isn't it possible to give an arbitrary length
> of data to fft like in numpy?

There is a mathematical background which is well explained here:

https://math.stackexchange.com/questions/77118/non-power-of-2-ffts#77152

May 18
On Thursday, 17 May 2018 at 13:36:39 UTC, kdevel wrote:
> On Thursday, 17 May 2018 at 12:34:25 UTC, Andre Pany wrote:
>> this applications throws an error in std.numeric (Line 2826).
>> => assert(isPowerOf2(range.length));
>
>> Isn't it possible to give an arbitrary length
>> of data to fft like in numpy?
>
> There is a mathematical background which is well explained here:
>
> https://math.stackexchange.com/questions/77118/non-power-of-2-ffts#77152

Thanks a lot for the details link.

Kind regards
André