Back to search

National and Kapodistrian University of Athens

Methodologies for accelerated analysis of the reliability and the energy efficiency levels of modern microprocessor architectures

Abstract

dc:description

The evolution in semiconductor manufacturing technology, computer architecture and design leads to increase in performance of modern microprocessors, which is also accompanied by increase in products’ vulnerability to errors. Designers apply different techniques throughout microprocessors life-time in order to ensure the high reliability requirements of the delivered products that are defined as their ability to avoid service failures that are more frequent and more severe than is acceptable.This thesis proposes novel methods to guarantee the high reliability and energy efficiency requirements of modern microprocessors that can be applied during the early design phase, the manufacturing phase or after the chips release to the market. The contributions of this thesis can be grouped in the two following categories according to the phase of the CPUs lifecycle that are applied at:• Early design phase: Statistical fault injection using microarchitectural structures modeled in performance simulators is a state-of-the-art method to accurately measure the reliability, but suffers from low simulation throughput. In this thesis, we firstly present a novel fully-automated versatile microarchitecture-level fault injection framework (called MaFIN) for accurate characterization of a wide range of hardware components of an x86-64 microarchitecture with respect to various fault models (transient, intermittent, permanent faults). Next, using the same tool and focusing on transient faults, we present several reliability and performance related studies that can assist design decision in the early design phases.Moreover, we propose two methodologies to accelerate the statistical fault injection campaigns. In the first one, we accelerate the fault injection campaigns after the actual injection of the faults in the simulated hardware structures. In the second, we further accelerate the microarchitecture level fault injection campaigns by proposing MeRLiN a fault pre-processing methodology that is based on the pruning of the initial fault list by grouping the faults in equivalent classes according to the instruction access patterns to hardware entries.• Manufacturing phase and release to the market: The contributions of this thesis in these phases of microprocessors life-cycle cover two important aspects. Firstly, using the 48-core Intel’s SCC architecture, we propose a technique to accelerate online error detection of permanent faults for many-core architectures by exploiting their high-speed message passing on-chip network. Secondly, we propose a comprehensive statistical analysis methodology to accurately predict at the system level the safe voltage operation margins of the ARMv8 cores of the X- Gene 2 chip when it operates in scaled voltage conditions.

Degree

thesis:*
Grantor dc:publisher
National and Kapodistrian University of Athens
Year dc:date
2018

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Kaliorakis, Emmanouil
  • Καληωράκης, Εμμανουήλ

Subjects

dc:subject × 22

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/42816

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Kaliorakis, Emmanouil; Καληωράκης, Εμμανουήλ. Methodologies for accelerated analysis of the reliability and the energy efficiency levels of modern microprocessor architectures. National and Kapodistrian University of Athens, 2018. http://hdl.handle.net/10442/hedi/42816