SM-centric transformation
Bo Wu, Guoyang Chen, Dong Li, Xipeng Shen, Jeffrey S. Vetter · 2014
To circumvent the limitation from the hardware scheduler on GPU, we create an SM-centric transformation technique. This technique enables complete control of the mapping between tasks and streaming multi-processors (SMs), and enables controlling the number of active thread blocks on each SM. Results show that our approach achieves better speedup than previous ones with kernel co-run cases.