Logo

Platform Lab Annual Retreat 2020 - Shared screen with speaker view
Ramki Krishnan
51:22
How do you see Network Slicing (across RAN etc.) playing a role in Private 5G especially given that RAN is a key bottle neck?
Deepak Vij (Futurewei)
01:28:42
Distributed Procedure Call — Collin LeeQuestion for Collin regarding DPC versus EDA (Event-Driven-Architecture). What you described, shouldn’t I be able to typically accomplish that using EDA (Fire&Forget model)? Just a curiosity.
Deepak Vij (Futurewei)
01:30:44
Collin, I will connect with you during the poster session to talk more on this. Quite interesting talk. Thanks.
Ramki Krishnan
01:47:04
Interesting talk on GMM – how much of this will be a real problem for single socket servers with single DRAM bank which is becoming a common deployment model?
Christos Kozyrakis
01:53:58
Hi Ramki, what do you mean single bank? Modern DDR3/4/5 is multibank to begin with. On top of that you have multiple channels even with a single CPU
Christos Kozyrakis
02:23:33
Seriously, AMD has been using NN-based branch prediction for a while now
Christos Kozyrakis
02:23:41
very shallow NNs
Christos Kozyrakis
02:23:49
(perceptrons)
Christos Kozyrakis
03:23:14
great talk Mark and Franky
Neeraja Yadwadkar (Stanford)
03:24:03
Yes, loved the demo!!
mark zhao
03:24:27
Thanks! :)