{"id":13933,"date":"2015-05-07T00:45:33","date_gmt":"2015-05-06T21:45:33","guid":{"rendered":"http:\/\/hgpu.org\/?p=13933"},"modified":"2015-05-07T00:45:33","modified_gmt":"2015-05-06T21:45:33","slug":"supporting-input-dependent-access-pattern-algorithms-on-gpus-using-gpufs","status":"publish","type":"post","link":"https:\/\/hgpu.org\/?p=13933","title":{"rendered":"Supporting input dependent access pattern algorithms on GPUs using GPUfs"},"content":{"rendered":"<p>Accelerating processing of very large datasets on GPUs is challenging, in particular when algorithms exhibit unpredictable data access patterns. In this paper we investigate the utility of GPUfs, a library that provides direct access to files from GPU programs, to implement such algorithms. We analyze the system&#8217;s bottlenecks, and suggest several modification to the GPUfs design, including new concurrent hash table for the buffer cache and a highly parallel memory allocator. We evaluate our changes by implementing a real image processing application which creates collages from a dataset of 2 Million images. The enhanced GPUfs design improves the application performance by 2x over the original GPUfs and outperforms both 12-core parallel CPU and standard CUDA-based GPU implementations, while significantly simplifying GPU application design.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Accelerating processing of very large datasets on GPUs is challenging, in particular when algorithms exhibit unpredictable data access patterns. In this paper we investigate the utility of GPUfs, a library that provides direct access to files from GPU programs, to implement such algorithms. We analyze the system&#8217;s bottlenecks, and suggest several modification to the GPUfs [&hellip;]<\/p>\n","protected":false},"author":351,"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":[36,89,33,3],"tags":[1787,14,1786,20,1470],"class_list":["post-13933","post","type-post","status-publish","format-standard","hentry","category-algorithms","category-nvidia-cuda","category-image-processing","category-paper","tag-algorithms","tag-cuda","tag-image-processing","tag-nvidia","tag-nvidia-geforce-gtx-titan"],"views":2204,"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/13933","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\/351"}],"replies":[{"embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=13933"}],"version-history":[{"count":0,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/13933\/revisions"}],"wp:attachment":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=13933"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=13933"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=13933"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}