Abstract:The PODEM (Path-Oriented Decision Making) algorithm is one of the classical path sensitive ATPG algorithms, which contains the backtrace process. Recent studies have shown that, in addition to the traditional heuristics, such as the SCOAP (Sandia Controllability and Observability Analysis Program) and COP (Controllability / Observability Procedure) measures, the ANN (Artificial neural network) can also guide the backtrace process in the PODEM algorithm. These methods are able to reduce the number of backtraces and the CPU time compared with the PODEM algorithms with the traditional heuristics. However, it may results in the negative improvements for some cases, due to the inherent errors in the output of ANN. To address this issue, an improved PODEM algorithm with the backtrace process guided by the incorporated ANN and traditional heuristics is proposed in this work. The effects of ANN hyper-parameters on the backtrace process of PODEM algorithm are studied, and the threshold which distinguishes the positive and negative improvements is discovered. The basic idea of the proposed PODEM algorithm is that, the traditional heuristics is applied to guide the backtrace process for each gate of the backtrace paths if the corresponding difference value of gate inputs in the ANN model is less than the threshold; and the ANN-based heuristics is applied vice versa. Evaluation results show that the proposed PODEM algorithm results in more positive improvements than the PODEM algorithm with the backtrace process guided by the ANN or the traditional heuristics alone. Furthermore, the proposed PODEM algorithm is better to be applied to the large size circuit, according to the application results on the benchmark circuits with different sizes.

A High Performance PODEM Algorithm with the Improved Backtrace Process

Improvement of random walk algorithm for power/ground grid analysis

A Provably Good and Practically Efficient Algorithm for Common Path Pessimism Removal in Large Designs

A Dedicated Greedy Short Path Padding Solution Method For Error Resilient Circuit Designs

A Two-Level Parallel Decomposition Approach for Transient Stability Constrained Optimal Power Flow

Improving branch divergence performance on GPGPU with a new PDOM stack and multi-level warp scheduling.

Parallel algorithm for edge tracing by use of GAPP

Power Gating Aware Task Scheduling in MPSoC

GPU-accelerated Path-based Timing Analysis.

POMP++: Facilitating Postmortem Program Diagnosis with Value-Set Analysis.

HeteroCPPR: Accelerating Common Path Pessimism Removal with Heterogeneous CPU-GPU Parallelism

Efficient Topology Reconfiguration for NoC-Based Multiprocessors: A Greedy-Memetic Algorithm

GPU-accelerated Critical Path Generation with Path Constraints

A memory-efficient neural ODE framework based on high-level adjoint differentiation

Performance-Aware Hybrid Algorithm For Mapping Ips Onto Mesh-Based Network On Chip

Postmortem Program Analysis with Hardware-Enhanced Post-Crash Artifacts.

PimPam: Efficient Graph Pattern Matching on Real Processing-in-Memory Hardware

Fork Path: Improving Efficiency Of Oram By Removing Redundant Memory Accesse

On-line MPSoC Scheduling Considering Power Gating Induced Power/Ground Noise

Improved pelican optimization algorithm with chaotic interference factor and elementary mathematical function

Improved MEOL and BEOL Parasitic-Aware Design Technology Co-Optimization for 3 Nm Gate-All-Around Nanosheet Transistor