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

Support for arbitrary strides #80

Closed
dtarakanov1 opened this issue Feb 18, 2019 · 2 comments
Closed

Support for arbitrary strides #80

dtarakanov1 opened this issue Feb 18, 2019 · 2 comments
Labels
enhancement request_for_comments See https://en.wikipedia.org/wiki/Request_for_Comments

Comments

@dtarakanov1
Copy link

MIOpen supports data with the NCHW layout. A tensor descriptor works dimensions in the NCHW order.

A tensor descriptor also requires an array of strides (see e.g. miopenSetTensorDescriptor()).

However according to the kernel code there is no such parameter as a height stride, it is always assumed to be equal to 1. This means that if input data is in the NHWC format, they must be physically re-ordered. It's not possible to simply change the strides order to get MIOpen algorithms working.

@daniellowell
Copy link
Contributor

@dtarakanov1 Thanks for your comment. Consistent striding is something we need to add across the library. Look for better support in the future releases.

@ppanchad-amd
Copy link

Apologies for the lack of response. Please try latest ROCm 6.0.2 to see if your issue has been resolved? If resolved, please close the ticket. Thanks.

@ppanchad-amd ppanchad-amd closed this as not planned Won't fix, can't repro, duplicate, stale Apr 2, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement request_for_comments See https://en.wikipedia.org/wiki/Request_for_Comments
Projects
None yet
Development

No branches or pull requests

5 participants