Don't have Telegram yet? Try it now!
https://hgpu.org/?p=9225
On the Efficacy of GPU-Integrated MPI for Scientific Applications