作者: Dirk Stroobandt , Ana Lucia Varbanescu , Catalin Bogdan Ciobanu , Muhammed Al Kadi , Andreas Brokalakis
DOI: 10.1109/RECOSOC.2016.7533896
关键词:
摘要: To handle the stringent performance requirements of future exascale-class applications, High Performance Computing (HPC) systems need ultra-efficient heterogeneous compute nodes. reduce power and increase performance, such nodes will require hardware accelerators with a high degree specialization. Ideally, dynamic reconfiguration be an intrinsic feature, so that specific HPC application features can optimally accelerated, even if they regularly change over time. In EXTRA project, we create new flexible exploration platform for developing reconfigurable architectures, design tools applications run-time built-in as core fundamental feature instead add-on. covers entire stack from architecture up to application, focusing on building blocks exascale systems: chip architectures very low overhead, truly take central concept, are tuned maximally benefit proposed techniques. Ultimately, this open improve Europe's competitive advantage leadership in field.