{"637425":{"#nid":"637425","#data":{"type":"event","title":"PhD Defense by Prasun Gera","body":[{"value":"\u003Cp\u003ETitle: Overcoming Memory Capacity Constraints for Large Graph Applications on GPUs\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nPrasun Gera\u003Cbr \/\u003E\r\nPh.D. candidate\u003Cbr \/\u003E\r\nSchool of Computer Science\u003Cbr \/\u003E\r\nGeorgia Institute of Technology\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nDate: Thursday, August 13th, 2020\u003Cbr \/\u003E\r\nTime: 3 p.m. - 5 p.m. (Eastern Time)\u003Cbr \/\u003E\r\nLocation: \u003Ca href=\u0022https:\/\/bluejeans.com\/392070473\u0022\u003Ehttps:\/\/bluejeans.com\/392070473\u003C\/a\u003E\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nCommittee:\u003Cbr \/\u003E\r\n---------------\u003Cbr \/\u003E\r\nDr. Hyesoon Kim (Advisor, School of Computer Science, Georgia Institute of Technology)\u003Cbr \/\u003E\r\nDr. Santosh Pande (School of Computer Science, Georgia Institute of Technology)\u003Cbr \/\u003E\r\nDr. Richard Vuduc (School of Computer Science and Engineering, Georgia Institute of Technology)\u003Cbr \/\u003E\r\nDr. Moinuddin Qureshi (School of Computer Science, Georgia Institute of Technology)\u003Cbr \/\u003E\r\nDr. Tushar Krishna (School of Electrical and Computer Engineering, Georgia Institute of Technology)\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nAbstract:\u003Cbr \/\u003E\r\n------------\u003Cbr \/\u003E\r\nGraphics Processing Units (GPUs) have been used successfully for accelerating a wide variety of\u003Cbr \/\u003E\r\napplications in the domains of scientific computing, machine learning, and data analytics over the\u003Cbr \/\u003E\r\nlast decade. Two important trends that have emerged across these domains are that for a lot of\u003Cbr \/\u003E\r\nreal-world problems, the working sets are larger than a GPU\u0026rsquo;s memory capacity, and that the data is\u003Cbr \/\u003E\r\nsparse. In this dissertation, we focus on graph analytics, for which the majority of prior work has\u003Cbr \/\u003E\r\nbeen restricted to graphs of modest sizes that fit in memory. Real world graphs such as social\u003Cbr \/\u003E\r\nnetworks and web graphs require tens to hundreds of gigabytes of storage whereas GPU memory is\u003Cbr \/\u003E\r\ntypically in the order of a few gigabytes. We investigate the following question: How can we\u003Cbr \/\u003E\r\naccelerate graph applications on GPUs when the graphs do not fit in memory?\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nThis question opens up two lines of inquiry. First, we consider the system architecture where the\u003Cbr \/\u003E\r\nGPU can address larger, albeit slower, host memory that is behind an interconnect such as PCI-e.\u003Cbr \/\u003E\r\nWhile this increases the total addressable memory, graph applications have poor locality that makes\u003Cbr \/\u003E\r\nefficient use of this architecture challenging. We formulate the locality problem as a graph\u003Cbr \/\u003E\r\nordering problem and\u0026nbsp;propose\u0026nbsp;efficient reordering methods for large graphs. The solution is general\u003Cbr \/\u003E\r\nenough that it can be extended to other similar architectures and even beneficial in cases where\u003Cbr \/\u003E\r\ngraphs fit in memory.\u003Cbr \/\u003E\r\n\u003Cbr \/\u003E\r\nSecond, we consider graph compression as a complementary approach. Conventional approaches to graph\u003Cbr \/\u003E\r\ncompression have been CPU-centric in that the decompression stage is sequential and branch\u003Cbr \/\u003E\r\nintensive. We devise an efficient graph compression format for large sparse graphs that is amenable\u003Cbr \/\u003E\r\nto run-time decompression on GPUs. The decompression stage is parallel and load balanced so that it\u003Cbr \/\u003E\r\ncan use the computational resources of GPUs effectively. In effect, analytics kernels can decompress\u003Cbr \/\u003E\r\nthe graph and do computation on the fly without decompressing the entire graph since the entire\u003Cbr \/\u003E\r\ndecompressed graph would not fit in memory.\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":" Overcoming Memory Capacity Constraints for Large Graph Applications on GPUs "}],"uid":"27707","created_gmt":"2020-07-31 17:08:21","changed_gmt":"2020-07-31 17:08:21","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2020-08-13T16:00:00-04:00","event_time_end":"2020-08-13T18:00:00-04:00","event_time_end_last":"2020-08-13T18:00:00-04:00","gmt_time_start":"2020-08-13 20:00:00","gmt_time_end":"2020-08-13 22:00:00","gmt_time_end_last":"2020-08-13 22:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"100811","name":"Phd Defense"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78761","name":"Faculty\/Staff"},{"id":"78771","name":"Public"},{"id":"78751","name":"Undergraduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}