Hybrid, scalable, trace-driven performance modeling of GPGPUs
Yehia Arafa, Abdel‐Hameed A. Badawy, Ammar ElWazir, Atanu Barai, Ali Eker, Gopinath Chennupati, Nandakishore Santhi, Stephan Eidenbenz
Abstract
In this paper, we present PPT-GPU, a scalable performance prediction toolkit for GPUs. PPT-GPU achieves scalability through a hybrid high-level modeling approach where some computations are extrapolated and multiple parts of the model are parallelized. The tool primary prediction models use pre-collected memory and instructions traces of the workloads to accurately capture the dynamic behavior of the kernels.
Topics & Concepts
Computer scienceScalabilityTRACE (psycholinguistics)Parallel computingCUDAComputationSupercomputerGeneral-purpose computing on graphics processing unitsComputer architectureComputational scienceProgramming languageOperating systemGraphicsPhilosophyLinguisticsParallel Computing and Optimization TechniquesAdvanced Data Storage TechnologiesDistributed and Parallel Computing Systems