Systems that comprise accelerators (e.g., GPUs) promise high performance, but their programming is still a challenge, mainly because of two reasons: 1) two distinct programming models have to be used within an application: one for the host CPU (e.g., C++), and one for the accelerator (e.g., OpenCL or CUDA); 2) using Just-In-Time (JIT) compilation and its optimization opportunities in both OpenCL and CUDA requires a cumbersome preparation of the source code. These two aspects currently lead to long, poorly structured, and error-prone GPU codes. Our PACXX programming approach addresses both aspects: 1) parallel programs are written using exclusively the C++ programming language, with modern C++14 features including variadic templates, generic lambda expressions, as well as STL containers and algorithms; 2) a simple yet powerful API (PACXX-Reflect) is offered for enabling JIT in GPU kernels; it uses lightweight runtime reflection to modify the kernel’s behaviour during runtime. We show that PACXX codes using the PACXX-Reflect are about 60% shorter than their OpenCL and CUDA Toolkit equivalents and outperform them by 5% on average.<\/p>\n","protected":false},"excerpt":{"rendered":"
Systems that comprise accelerators (e.g., GPUs) promise high performance, but their programming is still a challenge, mainly because of two reasons: 1) two distinct programming models have to be used within an application: one for the host CPU (e.g., C++), and one for the accelerator (e.g., OpenCL or CUDA); 2) using Just-In-Time (JIT) compilation and […]<\/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":[11,89,90,3],"tags":[955,1782,14,1483,1814,20,1793,193,1390],"class_list":["post-15445","post","type-post","status-publish","format-standard","hentry","category-computer-science","category-nvidia-cuda","category-opencl","category-paper","tag-compilers","tag-computer-science","tag-cuda","tag-intel-xeon-phi","tag-llvm","tag-nvidia","tag-opencl","tag-ptx","tag-tesla-k20"],"views":2553,"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/15445","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=15445"}],"version-history":[{"count":0,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/15445\/revisions"}],"wp:attachment":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=15445"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=15445"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=15445"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}