OSCAR Parallelizing and Power Reducing Compiler and API for Heterogeneous Multicores

Workshop:PEHC: Programming Environments for Heterogeneous Computing

Authors: Hironori Kasahara (Waseda University, Japan)

Abstract: This paper introduces the OSCAR Parallelizing and Power Reducing Compiler and API for shared memory heterogeneous multicore processors having general-purpose and accelerator cores. They parallelize and vectorize a sequential C or Fortran program automatically using Multigrain Parallelization, Data Locality Management for Cache, Distributed Shared Memory and Local Memory, Data Overlapping Transfer using Data Transfer Unit, Power Reducing using DVFS and Clock- & Power Gating, and so on.

Website:

Back to PEHC: Programming Environments for Heterogeneous Computing Archive Listing

Back to Full Workshop Archive Listing