PasTiS: building an NVIDIA Pascal GPU simulator for embedded AI applications

Michaël Adalbert (1, 2, 3) , Thomas Carle (2, 3) , Christine Rochange (2, 3)

1 IRT SystemX
2 IRIT-TRACES - Groupe de Recherche en Architecture et Compilation pour les systèmes embarqués
3 UT3 - Université Toulouse III - Paul Sabatier

We present PasTiS, a simulator for the NVIDIA Pascal GPU architecture family, with a focus on timing simulation. PasTiS supports a subset of the Pascal ISA, sufficient to simulate the execution of neural networks. We present this subset, as well as the underlying microarchitecture that we modelled using information available from NVIDIA, from scientific publications, and from our own experiments. We demonstrate the precision of the simulator by comparing it to measurements on the NVIDIA Jetson TX2 development board, on neural network applications.

Keywords: GPU; cycle-accurate simulator; timing analysis

Domains: Computer Science [cs]

Michaël Adalbert, Thomas Carle, Christine Rochange. PasTiS: building an NVIDIA Pascal GPU simulator for embedded AI applications. 11th European Congress on Embedded Real-Time Systems (ERTS 2022), 3AF Midi-Pyrénées: the French Society of Aeronautic and Aerospace; SEE : the French Society for Electricity, Electronics, and Information & Communication Technologies, Jun 2022, Toulouse, France. ⟨hal-03684680⟩