Synthesizing Code for GPGPUs from Abstract Formal Models
Speaker: Gabriel Hjort Blindell
This talk is about how to synthesize high-level, abstract formal models into code to be executed on GPGPUs. GPGPUs are massively data-parallel, through-put oriented platforms that can yield tremendous speedup of data-parallel applications, but they are also notoriously difficult to problem. We present a proof-of-concept synthesis tool - f2cc - which takes ForSyDe models as input and produces C + CUDA C code, the latter which is a language by NVIDIA for programming CUDA-based GPGPUs.