Software and Downloads

MAFIA: Multiple Application Framework for GPU Architectures (github)
MAFIA is developed for supporting multiple applications execution on GPUs. Currently, it supports 25 benchmarks from various benchmark suites (e.g., CUDA, Parboil, SHOC and Rodinia). From these benchmarks, one can construct 300 2-application workloads and 2300 3-application workloads.

Please cite the following paper if you use or build upon this framework:
(MEMSYS 2015) [PDF] [Talk (PPTX)] [BibTeX] [Github]
Adwait Jog, Onur Kayiran, Tuba Kesten, Ashutosh Pattnaik, Evgeny Bolotin, Niladrish Chatterjee, Stephen W. Keckler, Mahmut T. Kandemir, Chita R. Das,
Anatomy of GPU Memory System for Multi-Application Execution,
In the Proceedings of 1st International Symposium on Memory Systems (MEMSYS), Washington, DC, Oct 2015



Mars on GPGPU-Sim (tarball)
Mars is a MapReduce framework on GPUs. This package contains MapReduce implementations of some applications that can be executed on GPGPU-Sim, a cycle accurate GPU simulator.

Please cite the following paper if you use or build upon this framework:
(ASPLOS 2013) [PDF] [2-page-summary (PDF)] [Talk (PPTX)] [BibTeX]
Adwait Jog, Onur Kayiran, Nachiappan CN, Asit K. Mishra, Mahmut T. Kandemir, Onur Mutlu, Ravishankar Iyer, Chita R. Das, OWL: Cooperative Thread Array Aware Scheduling Techniques for Improving GPGPU performance, In the Proceedings of 18th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), Houston, TX, March 2013