Overview

Training Highlights

Intel/ALCF Marketing Highlights

oneAPI Ecosystem Highlights

COE Highlights

Training Highlights

James Reinders kicked off his XPU “How To” series with a blog Hello SYCL and DPC++ and three videos published on the Intel Software YouTube Channel:

• Why We care about XPUs (and What XPUs Are)
• Compile Hello SYCL to Learn SYCL with DPC++ on DevCloud
• Use the Intel® DevCloud

Intel® Hands-On Practice

• Get hands-on practice with these guided Jupyter* Notebooks on Data Parallel C++, OpenMP* with GPU Offload, Intel® oneAPI Math Kernel Library, and Intel® OSPRay Essentials  

ALCF Workshops

• April 20, 27-29 – GPU Hackathon (ALCF/NVIDIA)
• April 30 – Getting started on ThetaGPU
• May 4-6 – ALCF Computational Performance Workshop (https://www.alcf.anl.gov/events/2021-alcf-computational-performance-workshop)

Intel/ALCF Marketing Highlights

The Cross-Architecture challenge

• On March 25, Intel announced the winners of the Great Cross-Architecture Challenge, a collaboration with Argonne National Laboratory and CERN openlab. The challenge attracted participants in 52 countries across five continents, generated 28 high quality applications and demonstrated the growing momentum of oneAPI’s open, cross-architecture, multi-vendor programming approach. Entrants used oneAPI and Data Parallel C++ (DPC++) to create applications in areas such as bioinformatics, cryptography, data analytics, education, financial services, genomics, healthcare, image processing, mathematics, molecular dynamics, particle physics and ray tracing. Intel PR published a newsbyte here
• Andrew Pastrello of Sydney, Australia, won a chance to work on a oneAPI-related project with Argonne National Laboratory by showing the ease of porting CUDA code and modifying the application using DPC++ and oneAPI tools to synthesize audio from gravitational waveforms produced by black hole simulations to be used in music production and astronomy education

oneAPI 2021.2 release (May 30)

• We’ll promote the release as a whole as well as focuses social, blogs, etc. around specific features as appropriate 

ALCF Promotions

• Porting a particle-in-cell code to exascale architectures: https://www.alcf.anl.gov/news/porting-particle-cell-code-exascale-architectures
• Preparing for Exascale: Aurora supercomputer to help scientists visualize the spread of cancer: https://www.alcf.anl.gov/news/preparing-exascale-aurora-supercomputer-help-scientists-visualize-spread-cancer
• Let’s Talk Exascale Code Development: HACC: https://www.exascaleproject.org/lets-talk-exascale-code-development-hacc/

oneAPI Ecosystem Highlights

oneAPI Developer Summit at IWOCL 2021 on April 26th ( 8:00 am – 5:00 pm Munich time): oneAPI Dev Summit

• The first 2021 oneAPI Developer Summit will take place in conjunction with IWOCL & SYCLcon. In this one-day virtual conference, you will hear from industry and academia speakers working on innovative cross-platform, multi-vendor architecture solutions developed on oneAPI. Registration is open and Intel has >360 non-Intel registrations. Presentations and tutorials include Multivendor SYCL update Codeplay, Intel and Xilinx, Dev Cloud Updates, CoE HipSYCL, oneDNN on ARM and Fujitsu, Univ of Oregon, SENAI-CIMATEC, University of Lisboa, Univ of PittsBerg(SHREC), Creative Solution Space, Celerity and oneAPI community contest winners
• In addition to the oneAPI Developer Summit we’re a gold sponsor of the IWOCL conference and will be widely promoting the SYCL-related content
• Additional events by the oneAPI Dev Summit @ISC on June 22nd/23rd, an APJ Dev Summit in India on September 15th and then the PRC Dev Summit on September 16/17

COE Highlights

Argonne/Intel-COE events for the ECP Annual Meeting (ECPAM 2021).

• Intel COE Tutorial on SYCL and DPC++, April 12^th, 2021
• The COE organized a day-long non-NDA DPC++ Tutorial held on Intel DevCloud. The event was well attended with > 80 people logged into the Zoom channel for this event. The tutorial started with basics of SYCL & DPC++, transitioned into more advanced material, included deep dive session on how to use VTune for DPC++ workloads on Intel GPUs, and culminated with how DPC++ can be used today on Intel FPGAs. Additionally, beyond slide decks, all the content presented, including the FPGA material, was made available to attendees via Jupyter notebooks.  The presenters were John Pennycook, Thomas Applencourt (ANL), Ramesh Peri, Benjamin Odom, and Adonay Berhe. Members of the Intel COE were available throughout the session to assist users
• Path to Exascale (BOF session)
• This one-hour BOF session featured case studies from the ECP-ST (Kokkos, HYPRE) and ECP-AD (LQCD, NWChemEx) programs. HYPRE and NWChemEx focused on SYCL/DPC++, while Kokkos and LQCD focused on OpenMP-offload, thereby providing excellent coverage of the two major programming models for Aurora. Abhishek Bagusetty (ANL) led the presentation and discussion for HYPRE and NWChemEx. Patrick Steinbrecher was the lead for Kokkos and LQCD. > 50 individuals were in attendance, audience participation was active, and all material shared was non-NDA