Section: Partnerships and Cooperations

International Initiatives

• Program: Cátedras Francesas UFMG

• Title: Compiler Support for emerging parallel architectures

• Inria principal investigator: Sylvain Collange

• International Partner (Institution - Laboratory - Researcher):

  • Universidade Federal de Minas Gerais (UFMG) - Computer Science Department - Fernando Pereira

• Duration: Sep 2014 - Dec 2014

• We propose . The project develop compilation techniques for code optimization to speedup applications that run in Graphics Processing Units (GPUs). The objective is to enable developers code high-performance programs in high-level languages, while taking maximum benefit from the hardware. In particular, we seek to alleviate control and memory divergence, which are important performance limiters specific to GPU architectures. For instance, the call fusion optimization factors out a common function call invoked from multiple independent conditional branches to enable the hardware to execute the function in SIMD mode regardless of branch divergence.