Is anyone ever acces to complex64 value in rawkernel

[sim Bec]

Hello everyone,

It's the first time I use this group.

I'm a frenche engeneer in a university laboratory.

I've already developped a soft for analysing hologram in C++/CUDA.

For researcher and student I'm developping a new code in Python/CUPY

I've already quite translating all the functions needed but I have some problem when I want to access imaginary and real part of some complex cupy array in a rawkernel.

Here is my functions and the related error:

@jit.rawkernel()

def clean_plan_cplx_device(d_plan_cplx, size_x, size_y, posX, posY, clean_radius_pix, replace_cplx_value):

    index = jit.blockIdx.x * jit.blockDim.x + jit.threadIdx.x

    sizeXY = size_x * size_y

    jj = cp.int32(index) // cp.int32(size_x)

    ii = cp.int32(index) - cp.int32(jj * size_x)

    if (ii < size_x and jj < size_y):   #calcul distance

        distance = cp.sqrt((posX - ii)**2 + (posY - jj)**2)

        cplx = cp.complex64(d_plan_cplx[ii, jj])

        mod = cp.sqrt(cp.real(cplx)**2 + cp.imag(cplx)**2)

    if (distance < clean_radius_pix):

        d_plan_cplx[ii, jj] = 0.0+0j

    else:

        d_plan_cplx[ii, jj] = mod + 0j

def clean_plan_cplx(d_plan_cplx, size_x, size_y, posX, posY, clean_radius_pix, replace_value):

    nthread = 1024

    nBlock = math.ceil(size_x * size_y // nthread)

    clean_plan_cplx_device[nBlock, nthread](d_plan_cplx, size_x, size_y, posX, posY, clean_radius_pix, replace_value)

TypeError: Invalid function call 'real'.

Is there a way to acces real and imaginary part in this kernel?

Thank you for your help

Best regards

Simon