pyopencl | OpenCL integration for Python , plus shiny features | GPU library

I want to populate an array with float4 type. I have no idea how to initialize the arrays with something else than zeros. I've tried variations of this, but this is what I've come with, that explains what I want to do:

I faced very strange behavior of OpenCL. I've linked a minimal code sample.

Starting from some random index (commonly 32-divisible) values is not written to array if I add one extra operation beforehand (g_idata[ai] = g_idata[ai-1]). Also notable that, i will get correct result if:

1. just read value, and writing a literal (see SHOW_BUG).
2. add if (ai >= n) g_idata[0]+=0; at beginning. see commented lines

tested on Intel and nvidia.

I can't undestand where I make mistake. Target - fill array with unique work id.

I found magic number - q = 16777216 = 1024 * 1024 * 16. If array length less than q, all good: check_array function returns True with params:

I am new to OpenCL and trying to get the information about installed platforms and devices in my Machine. There is only a single platform installed in my PC and OpenCL is easily detecting it. The following C code:

I have a pyopencl based code which runs perfectly fine for 3-dimensional work groups, but when moving to 4-dimensional work groups, it breaks down with the error:

pyopencl._cl.LogicError: clEnqueueNDRangeKernel failed: INVALID_WORK_DIMENSION

Digging around, I found this answer to another question, which implies that OpenCl in fact allows higher dimensional work groups.

So my question is if it is possible to change this setting in pyopencl. From this other answer elsewhere, I understand that pyopencl immediately inputs the dimensions, but given the error I have, I think there must be some issue.

This is a minimal sample code to replicate this error. The code works well for the first kernel function, it breaks down on the second one.

I am trying to implement a bincount operation in OpenCL which allocates an output buffer and uses indices from x to accumulate some weights at the same index (assume that num_bins == max(x)). This is equivalent to the following python code:

I work Interchangeably with 32 bit floats and 32 bit integers. I want two kernels that do exactly the same thing, but one is for integers and one is for floats. At first I thought I could use templates or something, but it does not seem possible to specify two kernels with the same name but different argument types?

I wrote some code to parallelize convolution with one filter in python on my gpu. I keep receiving this error and I am unsure how to fix it. I posted the error below as well as my code. Thank you so much in advance.

I checked out some past stack overflow responses for this question but none of them seemed to do the trick. So it's possible I may have unaccounted for something that you may be able to catch.

I am having trouble understanding what the work item constraints mean. I am using pyopencl and looking at the max_work_item_sizes it gives what I assumed was the max number of global work threads for each dimension.