You might try loopy <> .  It allows out of the 
box execution of pyopencl kernels (assuming you have opencl correctly 
installed, etc.), with a simpler pseudo-code-esque syntax.  It sounds more 
along the lines of what you were looking for





From: PyOpenCL [] On Behalf Of Stephane 
Sent: Thursday, August 10, 2017 5:19 AM
Subject: Re: [PyOpenCL] getting started with pyopencl



thank you,

So, I will start softly with opencl…


I was hoping I could avoid the opencl/c++ step. that’s usually a big python 
advantage. I don’t have any clear difficulty yet, I was just planning to 
introduce some concepts and syntaxes and play with some codes (I have several 
Markov Chains - pymc3 codes for instance) and then debug a more elaborate 

for example I’ve seen there are several ways for opening a context while I 
don’t even know why having context is required...



On 10 août 2017, at 11:00, Sven Warris < <> 
> wrote:


Dear Stephane,

Starting out with parallel computing can be indeed a big challenge. 

On 10-08-17 10:10, Stephane Paulin-Henriksson wrote:

Unfortunately, I find the on-line documentation concerning pyopencl 
surprinsigly murky. I’m uncomfortable with it and need infos discussing both 
basics as well as more complex concepts.

Could you elaborate on the type of issues you have? Are you having difficulties 
with OpenCl and developing algorithms for parallel processing? Or is it the 
python-opencl interface?

What would be your advise to get started with pyopencl ?

I would suggest to start with a small example in OpenCL and then integrate it 
in python using pyOpenCL. 

See for example:

Good luck,



PyOpenCL mailing list

Reply via email to