{"id":15742,"date":"2016-04-19T23:09:48","date_gmt":"2016-04-19T20:09:48","guid":{"rendered":"http:\/\/hgpu.org\/?p=15742"},"modified":"2016-04-19T23:09:48","modified_gmt":"2016-04-19T20:09:48","slug":"a-unified-hardware-fitted-cross-gpu-performance-model","status":"publish","type":"post","link":"https:\/\/hgpu.org\/?p=15742","title":{"rendered":"A Unified, Hardware-Fitted, Cross-GPU Performance Model"},"content":{"rendered":"<p>We present a mechanism to symbolically gather performance-relevant operation counts from numerically-oriented subprograms (&#8216;kernels&#8217;) expressed in the Loopy programming system, and apply these counts in a simple, linear model of kernel run time. We use a series of &#8216;performance-instructive&#8217; kernels to fit the parameters of a unified model to the performance characteristics of GPU hardware from multiple hardware generations and vendors. We evaluate the predictive power of the model on a broad array of computational kernels relevant to scientific computing. In terms of the geometric mean, our simple, vendor- and GPU-type-independent model achieves relative accuracy comparable to that of previously published work using hardware specific models.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>We present a mechanism to symbolically gather performance-relevant operation counts from numerically-oriented subprograms (&#8216;kernels&#8217;) expressed in the Loopy programming system, and apply these counts in a simple, linear model of kernel run time. We use a series of &#8216;performance-instructive&#8217; kernels to fit the parameters of a unified model to the performance characteristics of GPU hardware [&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":[11,90,3],"tags":[1872,7,1782,20,1767,1793,67,1006,1543],"class_list":["post-15742","post","type-post","status-publish","format-standard","hentry","category-computer-science","category-opencl","category-paper","tag-amd-radeon-r9-fury","tag-ati","tag-computer-science","tag-nvidia","tag-nvidia-geforce-gtx-titan-x","tag-opencl","tag-performance","tag-tesla-c2070","tag-tesla-k40"],"views":2110,"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/15742","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=15742"}],"version-history":[{"count":0,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/15742\/revisions"}],"wp:attachment":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=15742"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=15742"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=15742"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}