Linear Algebra Algorithms for Hybrid Architectures with XKaapi
Instituto de Informatica – UFRGS, Grupo de Processamento Paralelo e Distribuido (GPPD)
IX Workshop of Parallel and Distributed Processing Group (WSPPD 2011), 2011
The emergence and continuing use of multicore architectures with GPU accelerators require changes in the current software to address the gap between the accelerators’ computer vs the CPU-GPU communication speed. We describe how to develop linear algebra algorithms for these new and emerging hybrid architectures using XKaapi.
January 14, 2012 by hgpu
