Please join us for the CUP-ECS Seminar, “DPC++: A modern C++ parallel programming model for Heterogeneous Computing,” presented by Nawrin Sultana of Intel Corporation.
Event Details: Friday, March 26th, 12 p.m. PST/1 p.m. MST/ 3 p.m. EST via Zoom. Contact Holley-Beeland@utc.edu for Zoom details.
Friday, March 26th, 3 p.m. EST
DPC++: A modern C++ parallel programming model for Heterogeneous Computing
Narwin Sultana, Intel Corporation
Abstract: Heterogeneous platforms are becoming increasingly common in HPC programming, with compute resources that include CPUs, GPUs, FPGAs, and other domain-specific compute engines. Understanding how to optimize computations for offload to these devices is an important and timely topic. The C++ language continues to grow in importance in High Performance Computing (HPC) programming. The Data Parallel C++ (DPC++) is based on industry-standard C++, incorporates SYCL specification from the Khronos Group, and includes language extensions developed using an open community process. It is designed as an open, cross-industry alternative to single-architecture, and proprietary languages. DPC++ enables developers to port their code more easily across CPUs, GPUs, and FPGAs, and also tune performance for a specific accelerator.
Bio: Dr. Sultana is a Software Engineer at Intel Corporation and started working there in 2020. She earned her Doctorate in Computer Science in 2019 and a Master of Science in Computer Science in 2016 from Auburn University. Her research interests include parallel and distributed architectures and applications; parallel programming paradigms; and fault tolerance at the application and system level. Her Ph.D. research included designing a fault-tolerant model for HPC applications that use Message Passing Interface (MPI). Currently, at Intel, Dr. Sultana is working on research and development of commercial and open-source parallel runtime libraries for multi-processor and heterogeneous architectures.