Skip to content

UCF Virtual Workshop 2020

Pavel Shamis (Pasha) edited this page Nov 7, 2020 · 55 revisions

Disclosure

You are registering for an open public standards setting discussion and development meeting of UCF. The discussions that take place during this meeting are intended to be open to the general public and all work product derived therefrom shall be made widely and freely available to the public. All information including exchange of technical information shall take place during open sessions of this meeting and UCF will not sponsor or support any private working group, standards setting or development sessions that may take place during this meeting. Your participation in any non-public interactions or settings during this meeting are outside the scope of UCF's intended open-public meeting format.

Dates

Nov 30 - Dec 3 (4 days)

Free registration for the event

Register here

Location

Virtual - ZOOM (Zoom link to be provided)

Agenda

Open UCX Agenda (USA Central time zone)

Date Time Topic Speaker/Moderator
11/30 08:00-09:00 UCF State of the Union Gilad Shainer, Nvidia & Pavel Shamis, Arm
09:00-10:00
GPU memory support

Seamless support of GPU buffers communication introduces yet another dimension for designing optimal data transfer protocols. In this talk we will have open discussion about outstanding issues and challenges, including out-of-box performance, PCI topology support, rendezvous protocols, memory type cache and memory hooks, and more.

Yossi Itigin, Nvidia

Yossi Itigin is a UCX team lead at NVIDIA, focuses on high-performance communication middleware, and a maintainer of OpenUCX project. Prior to joining NVIDIA, Mr. Itigin spent nine years at Mellanox Technologies in different technical roles, all related to developing and optimizing RDMA software.

10:00-11:00 ROCM + UCX Plans Sourav Chandra, AMD
11:00-12:00 UCX for Apache Spark Yossi Itigin, Nvidia
12:00-13:00 UCX Python - Dask/RAPIDS Ben Zaitlen, NVIDIA, Peter Entschev, NVIDIA, Matthew Baker, ORNL
12/01 08:00-09:00 UCF - Future directions Steve Poole, Los Alamos National Laboratory
09:00-10:00
UCP Protocols v2

In order to achieve out-of-box performance, UCX must select the optimal protocol for the given scenario, considering factors such as buffer length, memory locality, and data layout. In this talk we will present and discuss the next version of protocol selection mechanism, and make sure it will be able to handle that task.

Yossi Itigin, Nvidia

Yossi Itigin is a UCX team lead at NVIDIA, focuses on high-performance communication middleware, and a maintainer of OpenUCX project. Prior to joining NVIDIA, Mr. Itigin spent nine years at Mellanox Technologies in different technical roles, all related to developing and optimizing RDMA software.

10:00-11:00 UCP Active messages API Yossi Itigin, Nvidia
11:00-12:00 UCX development in Huawei Alex Margolin, Huawei
12:00-13:00 Open Smart NIC API - State of the Union Steve Poole, Los Alamos National Laboratory
12/02 08:00-09:00 BlazingSQL with UCX Rodrigo Aramburu & Felipe Aramburu, BlazingSQL
09:00-10:00 Charm++ with UCX Nitin Bhat, Charmworks
10:00-10:30 MPICH/UCX Update Ken Raffenetti ,Argonne National Laboratory
10:40-11:40 UCX counters in Score-P and Vampir Shuki Zanyovka, Huawei
11:40-12:40 Unified Communication Datatypes - State of the Union Pavan Balaji, Argonne National Laboratory
12:40-13:00
Arm IP building blocks and standards for SmartNIC

This talk describes the building-blocks available from Arm for modern SmartNICs/DPUs along with relevant standards enabling operating systems to boot on Arm-based SmartNICs without modification. Support for interfaces like CXL in Arm IP that targets future SmartNIC architectures is also discussed.

Kshitij Sudan, Arm

Kshitij the Technical Assistant to the GM of Arm’s Infrastructure business unit. He is a systems architect and focuses on enabling Arm customers targeted different market segments using Arm IP. Kshitij holds a PhD in computer architecture from University of Utah.

12/03 08:00-09:00 UCC: Design and Implementation of Next Generation Collectives Library Manju Gorentla, Nvidia
09:00-09:30 One-to-many UCT transports, part I: Shared-memory Alex Margolin, Huawei
09:30-10:00 One-to-many UCT transports, part II: Multicast Morad Horany, Huawei
10:00-11:00 Until UCC is available - UCG status update Alex Margolin, Huawei
11:00-11:45 RDMA-CORE Linux kernel and user space updates Jason Gunthorpe, Nvidia
11:45-12:30 RDMA-CORE DMABUF Jianxin Xiong, Intel
12:30-13:30 Scaling Facebook's Deep Learning Recommender Model (DLRM) with UCC/XCCL Josh Ladd, Nvidia & Srinivas, Facebook
13:30-14:00 Open Smart NIC API - OpenSHMEM I/O Extensions for Fine-grained Access to Persistent Memory Storage Megan Grodowitz, Arm
Clone this wiki locally