HarryWei / gloop

GLoop: An Event-driven Runtime for Consolidating GPGPU Applications

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

GLoop: An Event-driven Runtime for Consolidating GPGPU Applications

Requirements

  • GCC 4.9 (You need to modify /usr/lib/nvidia-cuda-toolkit/{gcc, g++})
  • NVIDIA CUDA 7.5 or later
  • grpc
  • CMake with CUDA patch (https://github.com/CPFL/cmake cuda branch)

Development

Do not use master branch, it is highly focusing on development purpose.

We have branches, kepler for Kepler K40c and pascal for Pascal P100.

Environment

  • Ubuntu 16.04

Publications

  • Yusuke Suzuki, Hiroshi Yamada, Shinpei Kato and Kenji Kono: GLoop: An Event-driven Runtime for Consolidating GPGPU Applications, In Proceedings of the 8th ACM Symposium on Cloud Computing (SoCC '17), 2017.
  • Yusuke Suzuki, Hiroshi Yamada, Shinpei Kato and Kenji Kono: Towards Multi-tenant GPGPU: Event-driven Programming Model for System-wide Scheduling on Shared GPUs, In The 2017 Workshop on Multicore and Rack-scale Systems (MaRS '16), 2016.

About

GLoop: An Event-driven Runtime for Consolidating GPGPU Applications


Languages

Language:Cuda 46.1%Language:C++ 26.6%Language:Perl 9.2%Language:C 7.3%Language:Makefile 4.6%Language:Ruby 2.5%Language:Python 1.4%Language:Shell 1.3%Language:CMake 0.9%Language:Assembly 0.1%Language:JavaScript 0.0%