reikna

Hi, sorry for a delayed response. How big is the array you're operating on? Can you show the full

Hi Bogan, Thank you so much, it is working!! El sábado, 20 de noviembre de 2021 a las 9:27:36 UTC+1,

Dear all, I have recently been trying Reikna to perform FFT and deconvolution algorithms using OpenCL

Thank you, I think I understand what's happening now. Transpose was trying to make a kernel with

> do they feedback of when they've sorted issues or is it more of we just try it after each

Hi Michael, Sorry for a delayed reply, I got caught up in work and completely forgot about this

I meant the problem size specifically, that is yes, the axes over which the transformation is

Thanks for all your help, I really appreciate it! With your changes I think I have a working version

Yes, it is intentional. Kernel.prepared_call() is low-level (CLUDA level), it does not know about

I think I realized the answer... Seems like computation's input is computation's input.

Hi Dmitry, You can use `functions.polar_unit()` (http://reikna.publicfields.net/en/latest/api/cluda.

It can be solved in momentum space, but that's not the method I'm using, because it cannot be

Thank you, I will take it into account then and only use power-of-2 sizes.

There are three possible ways. First, you can select the device manually and pass it to the Thread

I read the PyCUDA documentation and learned the CUDA contexts are a per-thread affair. I used that

Hi! Thanks for that tip. I'm using linux, where the multiprocessing module forks per default.

Source modules are compiled and executed on a GPU, so Python is not available there. You will have to

Hi Seth, I replied to you on SO, it is quite a long answer because I also had some comments about the

Hi Bogdan, It works! Thanks for bearing with me, Seth

Oh, I think I see what you mean. You can make it work by using `dtypes.c_constant()` function like so

Hi Florian, The thing with transformations is that they are called (in chain, according to their

Thanks for your answer Bogdan, it was very clear! Indeed, misc.lena() returns a gray level image.

Thanks sorted now.

No, unfortunately, currently there is no support in Reikna for that. In fact, there is not any

I've found the error in my code. All problem is solved. Thank you!

context.pop() is called when the Thread object is deleted, which in CPython will happen when the last

Thank you, i'm stick to use real input. I will follow you're first advice. Thanks.

if the result of fft using reikna is same as np.fft.ffn , the problem is solved. Thank you very much,

Ok, I submitted the bug, but it may be a while until there's any response. As for the compilation

The support for strides in PyOpenCL/PyCUDA was patchy when I last tried it, and I haven't