May 2019
Intermediate to advanced
452 pages
12h 16m
English
GPU devices with a compute capability of 3.5 or higher can make use of a feature that is primarily about reducing CPU-GPU interaction and instead works directly from the GPU. With dynamic parallelism, during runtime execution, threads can decide to launch configurations on the GPU itself. So, the need to transfer execution control and data between CPU and GPU can be greatly minimized.
Read now
Unlock full access