Ray tracing is an important and widely used tool in computer graphic. Entertainment and game industry have already benefit a lot from ray tracing. However, designers and end-users are forced to use off-line ray tracing tools for a long time due to the high computation load. In ray tracing, most of the computation is concentrated on whether hundreds of millions of rays hit objects in a scene. The naive algorithm calculate intersection between every ray and every triangle in the scene, which yields O(nm) computation time with n rays and m triangles. Inspired by binary search in one dimension space, one of the most popular and efficient improvement is partition the scene and build a tree to represent it. With this tree we have algorithm with O(nlog(m)). Travelling kd-tree with n rays is a natural parallel problem with each ray independent to each other. Thus if we have a kd-tree for a scene, the intersection problem can be calculated with O(log(m)) on n processors, which is an optimal algorithm with nm=log(m) accelerator. In practice, the accelerator is much lower due to hardware limitation, which we will discuss later. In this case kd-tree(k-dimensional tree)is developed with many other spacial partition algorithms. Despite animation with dynamic scene, kd-tree has been proved to be the most efficient data structure for static scene[2][16]. There are two tasks for us to use kd-tree in ray tracing: 1.Build kd-tree with triangle soup; 2.Travel kd-tree to find intersection points. This paper will first review state-of-the-art algorithms to build and travel kd-tree, both serial and parallel. Then implement an algorithm to build kd-tree with CUDA. Finally analyse this algorithm and try to make it faster and more memory efficient.<\/p>\n","protected":false},"excerpt":{"rendered":"
Ray tracing is an important and widely used tool in computer graphic. Entertainment and game industry have already benefit a lot from ray tracing. However, designers and end-users are forced to use off-line ray tracing tools for a long time due to the high computation load. In ray tracing, most of the computation is concentrated […]<\/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":[180,11,89,3],"tags":[1797,1782,14,412,20,1601,176,181],"class_list":["post-12054","post","type-post","status-publish","format-standard","hentry","category-3d-graphics-and-realism","category-computer-science","category-nvidia-cuda","category-paper","tag-3d-graphics-and-realism","tag-computer-science","tag-cuda","tag-kd-tree","tag-nvidia","tag-nvidia-geforce-gtx-780-ti","tag-package","tag-raytracing"],"views":6347,"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/12054","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=12054"}],"version-history":[{"count":0,"href":"https:\/\/hgpu.org\/index.php?rest_route=\/wp\/v2\/posts\/12054\/revisions"}],"wp:attachment":[{"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=12054"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=12054"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hgpu.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=12054"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}