Improving GPGPU Concurrency with Elastic Kernels