The Kokkos Lectures: Module 8 Q&A

12:10:48 From Milos Gligoric : sorry if I missed this; how did we figure out that something is wrong in the output? By looking at those numbers? 12:11:44 From Damien Lebrun-Grandie : Christian told us that the code usually can process 500x more particles per sec 12:12:09 From Milos Gligoric : ok thanks; missed that 13:02:12 From Anders Johansson : Are there any plans for functionality like tensordot/einsum? (For dense views) 13:05:16 From Siva Rajamanickam : We are starting to get more tensor users and have had requests for this. 13:05:31 From Siva Rajamanickam : I am assuming this 3d or more views. 13:05:57 From Anders Johansson : Ok, thanks! 13:06:05 From Anders Johansson : 5d in my case 13:06:24 From Junchao Zhang : What does the static mean in StaticCrsGraph? 13:06:40 From Balint Joo : Thanks everyone. I have to step away now. Congratulations on the Lecture Series and thank you for all the hard work you put into it. 13:06:43 From Brian Kelley : The number of nonzero per row is immutable 13:06:55 From Siva Rajamanickam : Thanks Balint ! 13:07:17 From Junchao Zhang : I see. I thought it was ‘static’ in C 13:08:35 From Siva Rajamanickam : The non-zero pattern of matrix is static. 13:08:58 From Siva Rajamanickam : This is to reuse it in simulations where structure remains the same. 13:08:59 From Junchao Zhang : Is SpMV optimized for alpha=1, beta=0? 13:11:02 From Seher Acer : Yes, it is 13:15:10 From Junchao Zhang : Do you support P^TAP? 13:15:39 From Luc Berger-Vergiat : Not directly 13:16:01 From Luc Berger-Vergiat : You can perform P^t AP using two calls to SpGEMM 13:16:47 From Junchao Zhang : I see. 13:17:54 From Siva Rajamanickam : Let me expand, this is mainly because most multigrid codes need communication between P^TA and multiply with P as of now. One could do all communication ahead of time, but that needs changes above KK. 13:37:06 From Seher Acer : Can everyone see the full slide? 13:37:18 From Siva Rajamanickam : Yes. 13:37:23 From Siva Rajamanickam : You can’t ? 13:37:23 From Luc Berger-Vergiat : Yes 13:37:34 From Seher Acer : I only see a portion of it 13:37:49 From Junchao Zhang : yes 13:37:56 From Siva Rajamanickam : Hope the recording is fine. 13:38:36 From Seher Acer : Oh I just changed the view option to “Fit to window” and it is fine now 13:38:42 From Siva Rajamanickam : :)