The exponential growth of floating point power in GPUs, combined with high
\nmemory bandwidth, has given rise to an attractive platform upon which to
\ndeploy HPC applications. We review the QUDA library which is a
\ndomain-specific library designed to accelerate legacy lattice quantum
\nchromodynamics application through providing a rich library of the common
\nperformance-critical algorithms, including highly optimized sparse linear
\nsolvers. QUDA is a toolbox in which to research and develop algorithms for
\nthe exascale future where flops are free and memory bandwidth determines
\nperformance. We describe the techniques currently present in QUDA and also
\nconsider future directions for development.<\/p>\n","protected":false},"featured_media":0,"template":"","tags":[],"seminar_category":[53],"acf":{"s_now_accepting":true,"s_date_order":"2013-07-01 10:00:00","s_date_end":null,"s_date_text":"","s_text":"Mike Clark","s_place":"other","s_place_other":"ES621","s_categoryother":"KMI Theory CTPLab Seminar","s_poster":"","s_poster2":"