Abstract to be presented at Summit 2016
High-performance computing (HPC) systems are now built around a de facto node architecture with latency-optimized multicore CPUs coupled to massively parallel bandwidth-optimized Accelerators. PGI Accelerator Fortran/C/C++ compilers, until now available exclusively on x86 CPU-based systems, focus on HPC optimizations and features such as automatic SIMD vectorization and support for high-level parallel and GPU programming. This talk will give an overview of status and early experiences with the forthcoming PGI compilers for OpenPOWER CPU-based systems. These compilers will provide HPC users with seamless migration of applications from Linux/x86+Tesla to Linux/OpenPOWER+Tesla using the same source code base, makefiles, build scripts and compiler commands and options. Included are examples of applications ported from Linux/x86 to Linux/OpenPOWER with a simple re-compile, and early performance results using CUDA Unified Memory to simplify GPU-enablement and optimization of HPC applications.
Doug Miles is director of PGI compilers & tools at NVIDIA. He has a background in mathematics, math library development, HPC benchmarking and applications. Doug has held a variety of applications engineering, technical marketing and management positions since joining PGI in 1993. He was an applications engineer at Cray Research Superservers and Floating Point Systems prior to joining PGI.