-
Notifications
You must be signed in to change notification settings - Fork 972
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
Fix scatter, reduce #1300
Fix scatter, reduce #1300
Conversation
2e51226
to
33b2b54
Compare
75d6e2b
to
dd3daf5
Compare
Ok I think this is ready for a review at least the reduction part (or I am just tired of improving it). I will add some benchmarks and then go ahead and benchmark the scatter that is also in this PR. It is a bit big and there are several changes on various parts so sorry about that. I will also write a long running test and a fuzzer to make sure that every size is properly checked in our nightly tests. All reduceWhen the array is big everything practically gets the memory bandwidth so it's all the same. But there is a difference for int8. Row reduceCol reduce |
🤤 |
I don't know what kind of magic you did here..but those graphs are pretty amazing. |
TODO: measure impact on runtime to see if we need to specialize this.