arXiv subjects

Jose Rodrigo Azambuja

Publications and source records attributed to Jose Rodrigo Azambuja.

5 recordsLinked to original sources

arXiv2026

SPINE: A Fault Injection Profiler for Quantized Neural Networks under Accumulated Faults

Deploying deep neural networks at the edge demands efficient inference under strict cost and power constraints. Quantized neural networks address these demands by replacing floating-point parameters with low-precision integers, yet their weights remain continuously exposed to radiation-induced bit-flips during inference. Fault Injection can be used to simulate those environments, but existing studies fail to characterize how accumulated upsets translate into mispredictions under realistic memory layouts. This paper presents a GDB-driven profiling framework that injects cumulative weight bit-flips directly onto the target binary of edge CPUs, generating per-layer fault profiles without requiring model retraining or code modification. Evaluated across multiple topologies, quantization efforts, and memory layouts, the results indicate how selective hardening strategies should be applied to effectively protect neural networks.

cs.AR↗

arXiv2026

A Tool for the Synthesis of Adaptive Probabilistic Processors Based on the Ising Model

This work presents a tool for the synthesis and simulation of probabilistic architectures for solving combinatorial optimization problems by mapping them to the Ising model. The proposed approach automatically constructs the Ising Hamiltonian and determines the number of probabilistic elements (p-bits) based on problem characteristics such as size and topology. Furthermore, the tool introduces an adaptive strategy for selecting the most suitable update algorithm among Gibbs Sampling, Simulated Annealing (SA), Simulated Quantum Annealing (SQA), and cluster-based methods. Experimental results using benchmark problems demonstrate improved convergence behavior and flexibility compared to fixed approaches. The proposed framework enables systematic evaluation of probabilistic computing strategies and supports the development of future hardware implementations based on MTJs and p-bits.

cs.AR↗

arXiv2024

NoX: a Compact Open-Source RISC-V Processor for Multi-Processor Systems-on-Chip

IoT applications are one of the driving forces in making systems energy and power-efficient, given their resource constraints. However, because of security, latency, and transmission, we advocate for local computing through multi-processor systems-on-chip (MPSoCs) for edge computing. The RISC-V ISA has grown in academia and industry due to its flexibility. Still, available open-source cores cannot be seamlessly integrated into MPSoCs for a fast time to market. This paper presents NoX, a compact open-source plug-and-play 32-bit RISC-V core designed in System Verilog for efficient data processing in MPSoCs. NoX has a 4-stage single-issue in-order pipeline with full bypass, providing an efficient resource-constrained architecture. Compared to industry and academia resource-constrained RISC-V cores, NoX offers a better resource usage and performance trade-off.

cs.AR↗

arXiv2023

Evaluating the Efficiency of Software-only Techniques to Detect SEU and SET in Microprocessors

This paper presents a detailed evaluation of the efficiency of software-only techniques to mitigate SEU and SET in microprocessors. A set of well-known rules is presented and implemented automatically to transform an unprotected program into a hardened one. SEU and SET are injected in all sensitive areas of a MIPS-based microprocessor architecture. The efficiency of each rule and a combination of them are tested. Experimental results show the inefficiency of the control-flow techniques in detecting the majority of SEU and SET faults. Three effects of the non-detected faults are explained. The conclusions can lead designers in developing more efficient techniques to detect these types of faults.

cs.AR↗

arXiv2022

Improving Content-Aware Video Streaming in Congested Networks with In-Network Computing

Network congestion and packet loss pose an ever-increasing challenge to video streaming. Despite the research efforts toward making video encoding schemes resilient to lossy network conditions, forwarding devices have not considered monitoring packet content to prioritize packets and minimize the impact of packet loss on video transmission. In this work, we advocate in favor of in-network computing employing a packet drop algorithm and an in-network hardware module to devise a solution for improving content-aware video streaming in congested network. Results show that our approach can reduce intra-predicted packet loss by over 80% at negligible resource usage and performance costs.

cs.NI↗