FATSEA - An Architectural Simulator for General Purpose Computing on GPUs

In this paper, the authors have presented a new framework for research on GPGPU devices from an architectural viewpoint, which opens the door for further research on architectural optimizations. Using a small set of benchmarks, they have implemented a well-known GPU architecture according to the freely available documentation from both industry and academia. They have also demonstrated that some platform details are still unknown under the hood: the simulated model is accurate when it comes to benchmarks requiring little memory overhead, but far from optimal when memory transfers predominate.

Provided by: University of Munster Topic: Hardware Date Added: Dec 2009 Format: PDF

Find By Topic