Enabling fast preemption via Dual-Kernel support on GPUs | IEEE Conference Publication | IEEE Xplore