-
Notifications
You must be signed in to change notification settings - Fork 218
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
Further unification of CUBLAS.axpy!
and LinearAlgebra.BLAS.axpy!
#432
Comments
I've been wary extending Instead, why not a method for |
Makes sense, I understand the concern now. My aim in using Julia's LinearAlgebra module does not document such an In any case, I agree this is a question for the Julia stdlib, rather than CUDA. Thank you. |
What about just using
Alternatively, we could implement |
Thanks for your help with this. I think my confusion arose from from the fact that |
The issue you filed ignores my question though: why can't you use |
Interesting idea. Here's what I've tried so far:
(v1) It's awesome that we can just write the naive broadcasting code, and it runs fast! (v2) I prefer the (v3) The final version with strided views works with CPU For my project, this feature isn't a blocker, obviously. But resolving the associated stdlib issue may eliminate a possible source of confusion. |
I am reopening, because on further reflection, I think @maleadt 's point is correct: Ideally, |
See #432. |
I'm interested in using
axpy!
with strides that differ from 1. Both BLAS and CUBLAS provide this functionality.For CPU arrays, the method is defined in
LinearAlgebra.BLAS.axpy!
hereFor CUDA arrays, the equivalent method is defined in
CUBLAS.axpy!
hereI'm wondering if the
LinearAlgebra.BLAS.axpy!
function should get methods for theCUBLAS.axpy!
behavior along these lines:The text was updated successfully, but these errors were encountered: