GPU Control Flow and Synchronization Slides 计算机科学

Lecture 3: control flow and synchronisationProf. Mike Gilesmike.giles@maths.ox.ac.ukOxford University Mathematical InstituteOxford e-Research CentreLecture 3 – p. 1Warp divergenceThreads are executed in warps of 32, with all threads in the warp executing the same instruction at the same timeWhat happens if different threads in a warp need to do different things?if (x