{"id":19047,"date":"2019-08-21T14:12:05","date_gmt":"2019-08-21T11:12:05","guid":{"rendered":"https:\/\/hgpu.org\/?p=19047"},"modified":"2019-08-25T15:27:30","modified_gmt":"2019-08-25T12:27:30","slug":"survey-paper-on-deep-learning-on-gpus","status":"publish","type":"post","link":"https:\/\/hgpu.org\/?p=19047","title":{"rendered":"Survey paper on Deep Learning on GPUs"},"content":{"rendered":"\n<p> The rise of deep-learning (DL) has been fuelled by the improvements in accelerators. GPU continues to remain the most widely used accelerator for DL applications. We present a survey of architecture and system-level techniques for optimizing DL applications on GPUs. We review 75+ techniques focused on both inference and training and for both single GPU and distributed system with multiple GPUs. It covers techniques for pruning, tiling, batching, impact of data-layouts, data-reuse schemes and convolution strategies (FFT\/direct\/GEMM\/Winograd), etc. It also covers techniques for offloading data to CPU memory for avoiding GPU-memory bottlenecks during training.&nbsp;<br> The paper is available&nbsp;<a rel=\"noreferrer noopener\" href=\"https:\/\/www.researchgate.net\/publication\/335292390_A_Survey_of_Techniques_for_Optimizing_Deep_Learning_on_GPUs\" target=\"_blank\">here<\/a>, accepted in J. of Systems Architecture 2019. <\/p>\n","protected":false},"excerpt":{"rendered":"<p>The rise of deep-learning (DL) has been fuelled by the improvements in accelerators. GPU continues to remain the most widely used accelerator for DL applications. We present a survey of architecture and system-level techniques for optimizing DL applications on GPUs. We review 75+ techniques focused on both inference and training and for both single GPU [&hellip;]<\/p>\n","protected":false},"author":615,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[3,1],"tags":[1673,1864,1867,1662],"class_list":["post-19047","post","type-post","status-publish","format-standard","hentry","category-paper","category-uncategorized","tag-deep-learning","tag-paper","tag-research","tag-survey"],"views":2827,"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/19047","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/users\/615"}],"replies":[{"embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=19047"}],"version-history":[{"count":4,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/19047\/revisions"}],"predecessor-version":[{"id":19059,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/19047\/revisions\/19059"}],"wp:attachment":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=19047"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=19047"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=19047"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}