HEP-CCE: Promoting Computational Excellence
HEP-CCE Coordinators: Salman Habib (Argonne), Kerstin Kleese Van Dam (Brookhaven), Rob Roser (Fermilab), and Peter Nugent (Lawrence Berkeley)
The HEP-CCE is a cross-cutting initiative to promote excellence in high performance computing (HPC) including data-intensive applications, scientific simulations, and data movement and storage. Enhancing connections with DOE’s Advanced Scientific Computing Research (ASCR) program is an important part of the Center’s activities. This includes promoting future-looking R&D initiatives in exascale architectures and systems, intelligent networking, and new data management and data analysis tools. Although the HEP-CCE is not a service-oriented entity, limited resources are available to support collaborative computing efforts for the HEP community, including a common GitHub repository for open source codes, a website for aggregating useful information, and expertise within and without HEP for solving computational problems via the Expert Forum. The HEP-CCE also sponsors topical workshops and student training programs.
Find out about our organization, projects, opportunities, and how you can become an HEP-CCE partner.
Applications Open for GPU Hackathon at Brookhaven National Laboratory, September 17-21, 2018
Brookhaven National Laboratory will host its 2nd GPU Hackathon on September 17-21, 2018. The Hackathon will bring together domain scientists, computational scientists, tools developers and hardware vendors to accelerate scientific codes on GPUs through the 5-day intense hands-on training. Team applications are being accepted until June 30, 2018. For more details, go tohttps://www.bnl.gov/gpuhackathon2018
INTRO TO HPC TRAINING (JUN 26-28)
The OLCF will host an Introduction to High Performance Computing (HPC) workshop on June 26-28, 2018. This training will start by covering basic skills, such as UNIX, vim text editor, and c/Fortran programming, which will be necessary for the topics to follow. We will then move on to cover basic parallel programming (using MPI and OpenMP) and GPU computing (CUDA and OpenACC). Hands-on sessions will be included with many of the topics to give participants the opportunity to practice new skills. For more information about this event or to register,
The OLCF will present an Introduction to Summit webinar from
1:00 PM until 4:30 PM (Eastern Time) on Friday, June 1. In this
webinar, we will cover the basic topics new users will need to
get up and running on Summit. We will give a broad overview of
available features and the details necessary to submit and run
jobs. For more information, please see the event page at
One challenge of using leadership computing resources is reading and writing data in scalable ways that does not lead to bottlenecks or performance penalties on the shared filesystems. This Workshop aims to bring together leading IO experts in the HEP field with experts from DOE ASCR Facilities to discuss how to move forward in the next years to make HEP software more friendly to millions of parallel threads accessing files on shared disks.
On February 26-March 2, 2018, Brookhaven National Laboratory (BNL) hosted a hackathon targeting the Intel Xeon Phi (formerly code-named Knights Landing, or KNL) processors for scientists to come together to optimize their application code performance on the KNL-based supercomputers. Out of the five teams who participated, two teams came from the high energy physics community:
- APES (Accelerator Particle Energy Simulator): A code for tracking particle-device and particle-particle interactions that has the potential to be used as the design platform for future particle accelerators. Team members all came from BNL.
- ART: a code for simulating the formation of structures in the universe, particularly galaxy clusters. Team members came from Yale University and University of Miami.
Each team was paired with a mentor with similar scientific background. They also had access to four floating mentors from Intel who brought expertise in OpenMP, Intel hardware architectures, compilers and performance profiling tools. The teams worked with their mentors for five days in a hands-on setting. By the end of the week, all teams achieved significant performance improvements for their codes, with ART and APES achieving >2X and >5X speedup, respectively.