GPU Accelerated Full System Simulation of Heterogeneous Many-core Platforms (SIMinG)

Students: Shivani Raghav

Contact person:  Dr. Martino Ruggiero, Prof. David Atienza

Partners

Electrical Engineering and Computer Science (DEIS), Bologna Univ., Italy;

Introduction

SIMinG is an open-source, fast, scalable and parallel simulator which is used for design space exploration and software development for future heterogeneous platforms with thousands of cores. It uses a novel mothodology of accelerating the simulation of many-core coprocessor using GPU platforms. The target architecture of the associated coprocessor consists of thousands of simple in-order cores connected in a tile network, Simulation of many target tile-thread is mapped to many hardware-threads that are available on high parallel GPU platforms. The full system simulator is partitioned between CPU and GPU, where the target general purpose CPU is simulated on host CPU using QEMU, whereas the many-core coprocessor is simulated on GPU platform.