Don't have Telegram yet? Try it now!
https://hgpu.org/?p=17860
Study of Bandwidth Partitioning for Co-executing GPU Kernels