Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Added NotImplemented error to Array.concatenate #508

Open
wants to merge 3 commits into
base: main
Choose a base branch
from

Conversation

panchoop
Copy link

@panchoop panchoop commented Aug 5, 2021

As suggested in this issue, a warning was added both in the code and in the documentation.

.. versionadded:: 2013.1
"""
# {{{ find properties of result array

shape = None
if axis != 0:
Copy link
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This isn't correct. This should instead be based on contiguity. Column-major arrays for example allow concatenation along the last axis.

Copy link
Author

@panchoop panchoop Aug 10, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I noticed that the code failed the tests because I missed a syntax error. Sorry for the noise.

About contiguity, I did some tests, and the function does not work in any axis for colum-major arrays. In line 2571 of the Array.concatenate function, the output array result is always a row-major array, leading to the same NotImlementedError due to miss-matching strides.

Simple example:

import pyopencl as cl
platform = cl.get_platforms()[0]
device = platform.get_devices()[0]
context = cl.create_some_context([device])
queue = cl.CommandQueue(context)

a = np.random.rand(2,2)
b = np.random.rand(2,2)

a_cl = clarray.to_device(queue, np.require(a, np.float64, 'F'))
b_cl = clarray.to_device(queue, np.require(b, np.float64, 'F'))

c_cl = clarray.concatenate( (a_cl, b_cl), axis=1)

It shouldn't be hard to allow column-major arrays, it is just a matter of initializing the result array with order="F" in these cases. The issue I see is that the Array class has no member that indicates the ordering. One could deduce the order by looking at the strides (if Array.strides[0] > Array.strides[-1]: row-major, else: column-major) but maybe it is too hacky. what do you think?

Another approach is the one taken in the Array.stack function, in which when the axis != 0, it is assumed that the input is column mayor. But it would also lead to unclear error messages.

Copy link
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for pointing out the issue with concatenate and column-major arrays, we should look into fixing that. I agree with your assessment about inspection of the strides, it's not quite elegant.

FWIW, CI is still not passing on this branch.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh yeah! I forgot to push. It should work now.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants