arXiv subjects

Saad Memon

Publications and source records attributed to Saad Memon.

2 recordsLinked to original sources

arXiv2026

Proton Irradiation Characterization of an Open-Source ML Accelerator on a Zynq UltraScale+ MPSoC

As spaceborne computing systems increasingly rely on neural network (NN) accelerators, the opacity of commercial, black-box architectures severely restricts the development of verifiable radiation mitigation strategies. Open-source, register-transfer level (RTL)-accessible accelerators resolve this limitation by enabling user-defined instrumentation, yet few have empirical radiation-response baselines. This work establishes a foundational system-level proton-irradiation baseline for an unmitigated open-source Tensil NN accelerator deployed on a Zynq UltraScale+ SoC executing ResNet-20 inference. Under 20 to 58 MeV proton irradiation, we delivered $4.29 \times 10^{10}$ p/cm$^{2}$ within monitored operational windows. Seven workload interruptions required two restarts of the notebook process, four reboots or board resets, and one power-cycle sequence. Two output-corruption events returned incorrect CIFAR-10 classes without loss of service. In the longer event, the accelerator returned a class absent from the ten-image CIFAR-10 pool for 39 consecutive inputs at normal cadence. The process remained alive, while the kernel log, limited memory test, and sampled power showed no anomaly. Observation of the stuck-class sequence ended with scheduled bitstream reconfiguration. All nine onsets occurred under the nominal 4 cm beam, which exposed the SoC, LPDDR4, and additional board circuitry; none occurred under the 2 cm SoC-centered field. This pattern shows a field association but does not establish LPDDR4 as the cause because field size was confounded with run order and dose. Linux-managed accelerators require end-to-end content checks and recovery that reaches the state in which corruption can persist. This baseline documents availability loss and silent output corruption, supporting future software hardening of COTS FPGA-SoCs for neural-network inference in space systems.

cs.AR↗

arXiv2025

Where Linux Breaks Under Radiation: A Cross-Architecture Kernel-Level Characterization of Proton-Induced Failures in COTS SoCs

Linux is increasingly deployed in Low Earth Orbit on commercial off the shelf systems on chip that were not designed for space radiation. Ionizing particles can trigger single event functional interrupts that crash the kernel without warning. Prior work mainly measured board level cross sections, leaving unclear which Linux subsystems fail and how a single upset propagates into an operating system wide failure across architectures, stress conditions, and irradiation conditions. We address this gap by subjecting three Linux platforms to proton irradiation in the 20 to 58 MeV range: a Raspberry Pi Zero 2W with a 40 nm planar ARM Cortex A53, an NXP i MX 8M Plus with a 14 nm FinFET ARM Cortex A53, and an OrangeCrab ECP5 FPGA hosting a VexRiscV RV32I soft core at 40 nm. Through kernel log forensics, we trace all 133 observed Linux failures, most of which have not been previously reported, to their originating kernel handlers. Failure profiles differ sharply across nodes. On the two 40 nm platforms, memory management and driver handlers account for 67 to 78% of events, while on the 14 nm SoC approximately 90% of failures funnel through a single eMMC storage path, comprising 56% filesystem failures and 34% driver failures. This shows that a SEFI susceptible peripheral can strongly dictate system reliability. The 14 nm SoC also shows roughly an order of magnitude lower Linux SEFI cross section, although irradiation geometry and DRAM exposure differences preclude isolating the contribution of process scaling. Reconstructed propagation chains show that faults can cascade through up to six kernel subsystems before terminal failure in severe events. Rather than motivating blanket redundancy, these results identify the kernel subsystem boundaries where radiation induced faults originate, enabling targeted mitigations for hardening COTS Linux systems for orbit.

cs.OS↗