%0 Conference Paper %F Oral %T PasTiS: building an NVIDIA Pascal GPU simulator for embedded AI applications %+ IRT SystemX %+ Groupe de Recherche en Architecture et Compilation pour les systèmes embarqués (IRIT-TRACES) %+ Université Toulouse III - Paul Sabatier (UT3) %A Adalbert, Michaël %A Carle, Thomas %A Rochange, Christine %< avec comité de lecture %Z confiance.ai %Z EC7 %B 11th European Congress on Embedded Real-Time Systems (ERTS 2022) %C Toulouse, France %8 2022-06-01 %D 2022 %K GPU %K cycle-accurate simulator %K timing analysis %Z Computer Science [cs]Conference papers %X 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. %G English %2 https://ut3-toulouseinp.hal.science/hal-03684680/document %2 https://ut3-toulouseinp.hal.science/hal-03684680/file/ERTS2_2022-9.pdf %L hal-03684680 %U https://ut3-toulouseinp.hal.science/hal-03684680 %~ UNIV-TLSE2 %~ UNIV-TLSE3 %~ CNRS %~ UT1-CAPITOLE %~ IRT-SYSTEMX %~ IRIT %~ IRIT-TRACES %~ IRIT-ASR %~ IRIT-UT3 %~ CONFIANCEAI %~ TOULOUSE-INP %~ ERTS2022 %~ UNIV-UT3 %~ UT3-INP %~ UT3-TOULOUSEINP