-
Notifications
You must be signed in to change notification settings - Fork 1
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
Refactor and add StridedView
support
#5
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In general, looks good to me!
I think there are quite a few tests that are not really using the backend, or might not make a lot of sense to test here, I think it is fine to assume that if the interface functions work, and a few tensor contractions too, everything should be fine.
One additional note is that for the tensor contractions, we might want to have some (more) tests with Bumper and the manual allocator, just to make sure that these are also supported.
Ok, I think I covered all of your comments. I just changed everything locally and pushed the changes, which was easier than committing some of the changes you made and merging them with the ones I already had locally. |
ef42e11
to
f63a457
Compare
f63a457
to
efc9fec
Compare
Complete overhaul to mimick the strided implementation from TensorOperations.jl. Also more tests.
It seems TBLIS does support a beta argument by encoding it in the C tensor structure, before entering the add or mult method. It also supports the conjugation flag in the add method (also used for trace), but unfortunately not in the mult (tensor contraction) method. We work around this by allocating a temporary for the conjugated tensor.
I think this is ready for review.