incorrect result for dpnp.fft.irfft

[vtavana]

The following example returns different results compare to numpy

import numpy
import dpnp
from numpy.random import random
from numpy.testing import assert_allclose

a = random((3,4)) + 1j * random((3,4))
ia = dpnp.array(a)
expected = numpy.fft.irfft(a, axis=1, n=8)
result = dpnp.fft.irfft(ia, axis=1, n=8)
assert_allclose(result.asnumpy(), expected)

#AssertionError: 
#Not equal to tolerance rtol=1e-07, atol=0

#Mismatched elements: 24 / 24 (100%)
#Max absolute difference among violations: 0.1234781
#Max relative difference among violations: 23.51308712
# ACTUAL: array([[ 0.606494, -0.30548 , -0.170699,  0.216856, -0.056749, -0.030368,
#        -0.268099,  0.244826],
#       [ 0.20851 , -0.108572, -0.199412,  0.064543,  0.120729,  0.298316,...
# DESIRED: array([[ 0.608355, -0.307341, -0.168838,  0.214995, -0.054888, -0.032229,
#        -0.266238,  0.242965],
#       [ 0.331988, -0.232051, -0.075934, -0.058936,  0.244207,  0.174838,...

[vtavana]

Input array for dpnp.fft.irfft should be Hermitian, otherwise the behavior is undefined.

The input in the example here is not Hermitian so failure is expected.