Files in this item

FilesDescriptionFormat

application/pdf

application/pdfUILU-ENG-17-2203.pdf (1MB)
(no description provided)PDF

Description

Title:Zoom Out and See Better: Scalable Message Tracing for Post-Silicon SoC Debug
Author(s):Pal, Debjit; Vasudevan, Shobha
Subject(s):Post-silicon debug
Observability
Tracing hardware
OpenSPARC T2
Abstract:We present a method for selecting trace messages for post-silicon validation of System-on-Chip (SoC). Our message selection is guided by specifications of interacting flows in common user applications. In current practice, such messages are selected based on designer expertise. We formulate the problem as an optimization of mutual information gain and trace buffer utilization. Our approach scales to systems far beyond the capacity of current signal selection techniques. We achieve an average trace buffer utilization of 98.96% with an average flow specification coverage of 94.3% and an average bug localization to only 21.11% of the potential root causes in our large-scale debugging effort. We present efficacy of our selected messages in debugging and root cause analysis using five realistic case studies consisting of complex and subtle bugs from the OpenSPARC T2 processor.
Issue Date:2017-11
Publisher:Coordinated Science Laboratory, University of Illinois at Urbana-Champaign
Series/Report:Coordinated Science Laboratory Report no. UILU-ENG-17-2203
Genre:Technical Report
Type:Text
Language:English
URI:http://hdl.handle.net/2142/98857
Sponsor:IBM
Date Available in IDEALS:2017-12-01


This item appears in the following Collection(s)

Item Statistics