Omit the alignment check when using MKL's fftw3 wrapper?

[N5N3]

As far as i know, the 'UNALIGNED' flag changes nothing when using MKL's fftw3 wrapper, as the low level DFTI API uses no pointer information until call the DftiComputeForward or DftiComputeBackward.
Some 1d-fft benchmark:

using FFTW, BenchmarkTools, LinearAlgebra
const memory = zeros(Int8,8*2048*20000 + 15) # memory pool for arbitrary alignment
pickalign(::Type{T}, sz, align::Int) where{T} = begin
    data = @view memory[align .+ (1:sizeof(T)*prod(sz))]
    dataᵀ = reinterpret(T,data)
    reshape(dataᵀ,sz)
end 
using Random
a = pickalign(ComplexF32,(2048,100), 0)
randn!(a)
b = similar(a)
plan = plan_fft(a,1)
plan_unaligned = plan_fft(a,1,flags = FFTW.ESTIMATE | FFTW.UNALIGNED)
@btime FFTW.mul!(b,plan,a)                     #fftw:58.600 μs      mkl::34.200 μs
@btime FFTW.mul!(b,plan_unaligned,a)   #fftw:153.800 μs     mkl::34.200 μs
a = pickalign(ComplexF32,(2048,100), 1)  
randn!(a)
@btime FFTW.mul!(b,plan_unaligned,a)  #fftw:152.700 μs      mkl::35.600 μs
a = pickalign(ComplexF32,(2048,100), 8)
randn!(a)
@btime FFTW.mul!(b,plan_unaligned,a)  #fftw:152.100 μs      mkl::35.700 μs

Although only a few cases have been benched, it seems that the alignment doesn't affect MKL's performance, and the default flags could be settled as ESTIMATE | UNALIGNED.

[N5N3]

Well I found this in MKL's developer reference [p2643]:
The flags parameter in plan creation functions is always ignored. The same algorithm is used regardless of the value of this parameter.