CUDA Portability with HIPCL and Intel® DPC++ Compatibility Tool

HIPCL expands the scope of the CUDA* portability route from an AMD platform to an OpenCL™ platform. The Intel® DPC++ Compatibility Tool migrates a CUDA program to a Data Parallel C++ (DPC++) program. With the goal of portability enhancement, we evaluate the performance of the CUDA applications from Rodinia, SHOC, and proxy applications ported using HIPCL and the Intel DPC++ Compatibility Tool on Intel GPUs. After profiling the ported programs, we look for performance gaps to optimize codes converted by the Intel DPC++ Compatibility Tool to improve their performance. The open source repository for the CUDA, HIP, and the Intel DPC++ Compatibility Tool programs will be useful for the development of a translator.

Speaker

Zheming Jin earned his PhD in computer science and engineering at the University of South Carolina where his research focused on program synthesis and application development targeting FPGAs. As a postdoctoral appointee at Argonne, he evaluated high-level synthesis and oneAPI on FPGAs and GPUs, respectively. His research focus is heterogeneous computing.