EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Exploiting mixed-precision redundancy for soft-error detection in LU decomposition

Nima Sahraneshinsamani, Sandra Catalán, José R. Herrero

The Journal of Supercomputing · 09.04.2026 · 82 (5)

DOI: 10.1007/s11227-026-08413-9 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of Supercomputing, online erschienen am 09.04.2026. Verfasst von Nima Sahraneshinsamani und 2 weiteren Autor:innen. Die Arbeit nennt 15 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Mixed-precision arithmetic, widely adopted for performance, creates an overlooked opportunity for fault tolerance. Our mixed-precision panel factorization () algorithm already computes each LU panel in both and ; we show that this inherent redundancy enables soft-error detection at negligible additional cost. We propose four complementary detectors that compare checksum sketches of the two factorizations: (thresholded checksum discrepancy), (outlier analysis of per-row differences), (a randomized-probe combination of both with separate $$L$$ L - and $$U$$ U -factor checks), and (a lightweight cross-check inspired by algorithm-based fault tolerance). Experimental evaluation shows that and achieve high sensitivity with low false-positive rates, while maintains near-zero false positives with moderate sensitivity and complementary coverage for challenging significand-bit faults. Bit-level analysis confirms reliable detection of exponent- and sign-bit errors, with sensitivity decreasing gracefully for low-order significand bits whose perturbations approach the rounding floor. A lightweight panel-fingerprint mechanism extends protection beyond the factorization loop, closing the temporal gap before the triangular solve with guaranteed detection of any single-element corruption and zero false positives. The approach requires only linear work and constant storage per panel, preserving the cubic scaling of standard LU decomposition.

Zitieren

APA

Sahraneshinsamani, N., Catalán, S., & Herrero, J. R. (2026). Exploiting mixed-precision redundancy for soft-error detection in LU decomposition. The Journal of Supercomputing, 82(5). https://doi.org/10.1007/s11227-026-08413-9

Vancouver

Sahraneshinsamani N, Catalán S, Herrero JR. Exploiting mixed-precision redundancy for soft-error detection in LU decomposition. The Journal of Supercomputing. 2026;82(5). doi:10.1007/s11227-026-08413-9

BibTeX
@article{Sahraneshinsamani2026,
  author = {Sahraneshinsamani, Nima and Catalán, Sandra and Herrero, José R.},
  title = {Exploiting mixed-precision redundancy for soft-error detection in LU decomposition},
  journal = {The Journal of Supercomputing},
  year = {2026},
  volume = {82},
  number = {5},
  doi = {10.1007/s11227-026-08413-9}
}
Kostenlos starten →