A Clang tool that transform any CUDA code to the SM-Centric form
[ICS'15] "Enabling and Exploiting Flexible Task Assignment on GPU through SM-Centric Program Transformations", Bo Wu, Guoyang Chen, Dong Li, Xipeng Shen, Jeffrey Vetter, ACM International Conference on Supercomputing, Newport Beach, CA, 2015. (25% acceptance rate) PDF