EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Fast and Accurate CPU Performance Modeling With Compositional Analytical–Machine Learning Fusion

Arash Nasr-Esfahany, Mohammad Alizadeh, Victor Lee, Hanna Alam, Brett W. Coon, David E. Culler, Vidushi Dadu, Martin Dixon, Henry M. Levy, Santosh Pandey, Parthasarathy Ranganathan, Amir Yazdanbakhsh

IEEE Micro · 2026-09 · 46 (5) · 29-36

DOI: 10.1109/mm.2026.3699340 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Micro, 2026-09. Verfasst von Arash Nasr-Esfahany und 11 weiteren Autor:innen. Die Arbeit nennt 9 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Cycle-level simulators, like gem5, are widely used in microarchitecture design, but prohibitively slow for large-scale design space exploration. We present Concorde, a new methodology for learning fast and accurate performance models of microarchitectures. Unlike existing simulators and learning approaches that emulate each instruction, Concorde uses simple analytical models to compute per component performance bounds, producing compact distributions that a machine learning (ML) model maps to program behavior. Concorde is more than five orders-of-magnitude faster than Google’s cycle-level simulator, with ∼2% average cycles-per-instruction (CPI) prediction error across SPEC2017, open source, and Google’s proprietary benchmarks. Concorde is already shaping large-scale architecture exploration at Google, contributing to internal CPU design projects. This enables analyses previously infeasible: In about an hour, we conducted a first-of-its-kind fine-grained performance attribution to microarchitectural components, requiring ∼150 M CPI evaluations. The same compositional analytical–ML recipe extends beyond CPUs, suggesting a broader approach to building fast, interpretable performance models.

Zitieren

APA

Nasr-Esfahany, A., Alizadeh, M., Lee, V., Alam, H., Coon, B. W., Culler, D. E., Dadu, V., Dixon, M., Levy, H. M., Pandey, S., Ranganathan, P., & Yazdanbakhsh, A. (2026). Fast and Accurate CPU Performance Modeling With Compositional Analytical–Machine Learning Fusion. IEEE Micro, 46(5), 29-36. https://doi.org/10.1109/mm.2026.3699340

Vancouver

Nasr-Esfahany A, Alizadeh M, Lee V, Alam H, Coon BW, Culler DE, et al. Fast and Accurate CPU Performance Modeling With Compositional Analytical–Machine Learning Fusion. IEEE Micro. 2026;46(5):29-36. doi:10.1109/mm.2026.3699340

BibTeX
@article{NasrEsfahany2026,
  author = {Nasr-Esfahany, Arash and Alizadeh, Mohammad and Lee, Victor and Alam, Hanna and Coon, Brett W. and Culler, David E. and Dadu, Vidushi and Dixon, Martin and Levy, Henry M. and Pandey, Santosh and Ranganathan, Parthasarathy and Yazdanbakhsh, Amir},
  title = {Fast and Accurate CPU Performance Modeling With Compositional Analytical–Machine Learning Fusion},
  journal = {IEEE Micro},
  year = {2026},
  volume = {46},
  number = {5},
  pages = {29--36},
  doi = {10.1109/mm.2026.3699340}
}
Kostenlos starten →