Source author record

Mircea R. Stan

Mircea R. Stan appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

cond-mat.mes-hall Hardware Architecture

Catalog footprint

What is connected

4works

2topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2016arXiv

CLEAR: Cross-Layer Exploration for Architecting Resilience - Combining Hardware and Software Techniques to Tolerate Soft Errors in Processor Cores

We present a first of its kind framework which overcomes a major challenge in the design of digital systems that are resilient to reliability failures: achieve desired resilience targets at minimal costs (energy, power, execution time, area) by combining resilience techniques across various layers of the system stack (circuit, logic, architecture, software, algorithm). This is also referred to as cross-layer resilience. In this paper, we focus on radiation-induced soft errors in processor cores. We address both single-event upsets (SEUs) and single-event multiple upsets (SEMUs) in terrestrial environments. Our framework automatically and systematically explores the large space of comprehensive resilience techniques and their combinations across various layers of the system stack (586 cross-layer combinations in this paper), derives cost-effective solutions that achieve resilience targets at minimal costs, and provides guidelines for the design of new resilience techniques. We demonstrate the practicality and effectiveness of our framework using two diverse designs: a simple, in-order processor core and a complex, out-of-order processor core. Our results demonstrate that a carefully optimized combination of circuit-level hardening, logic-level parity checking, and micro-architectural recovery provides a highly cost-effective soft error resilience solution for general-purpose processor cores. For example, a 50x improvement in silent data corruption rate is achieved at only 2.1% energy cost for an out-of-order core (6.1% for an in-order core) with no speed impact. However, selective circuit-level hardening alone, guided by a thorough analysis of the effects of soft errors on application benchmarks, provides a cost-effective soft error resilience solution as well (with ~1% additional energy cost for a 50x improvement in silent data corruption rate).

preprint2013arXiv

Computing with Non-equilibrium Ratchets

Electronic ratchets transduce local spatial asymmetries into directed currents in the absence of a global drain bias, by rectifying temporal signals that reside far from thermal equilibrium. We show that the absence of a drain bias can provide distinct energy advantages for computation, specifically, reducing static dissipation in a logic circuit. Since the ratchet functions as a gate voltage-controlled current source, it also potentially reduces the dynamic dissipation associated with charging/discharging capacitors. In addition, the unique charging mechanism eliminates timing related constraints on logic inputs, in principle allowing for adiabatic charging. We calculate the ratchet currents in classical and quantum limits, and show how a sequence of ratchets can be cascaded to realize universal Boolean logic.

preprint2011arXiv

Graphene Nanoribbons: from chemistry to circuits

The Y-chart is a powerful tool for understanding the relationship between various views (behavioral, structural, physical) of a system, at different levels of abstraction, from high-level, architecture and circuits, to low-level, devices and materials. We thus use the Y-chart adapted for graphene to guide the chapter and explore the relationship among the various views and levels of abstraction. We start with the innermost level, namely, the structural and chemical view. The edge chemistry of patterned graphene nanoribbons (GNR) lies intermediate between graphene and benzene, and the corresponding strain lifts the degeneracy that otherwise promotes metallicity in bulk graphene. At the same time, roughness at the edges washes out chiral signatures, making the nanoribbon width the principal arbiter of metallicity. The width-dependent conductivity allows the design of a monolithically patterned wide-narrow-wide all graphene interconnect-channel heterostructure. In a three-terminal incarnation, this geometry exhibits superior electrostatics, a correspondingly benign short-channel effect and a reduction in the contact Schottky barrier through covalent bonding. However, the small bandgaps make the devices transparent to band-to-band tunneling. Increasing the gap with width confinement (or other ways to break the sublattice symmetry) is projected to reduce the mobility even for very pure samples, through a fundamental asymptotic constraint on the bandstructure. An analogous trade-off, ultimately between error rate (reliability) and delay (switching speed) can be projected to persist for all graphitic derivatives. Proceeding thus to a higher level, a compact model is presented to capture the complex nanoribbon circuits, culminating in inverter characteristics, design metrics and layout diagrams.

preprint2010arXiv

Monolithically Patterned Wide-Narrow-Wide All-Graphene Devices

We investigate theoretically the performance advantages of all-graphene nanoribbon field-effect transistors (GNRFETs) whose channel and source/drain (contact) regions are patterned monolithically from a two-dimensional single sheet of graphene. In our simulated devices, the source/drain and interconnect regions are composed of wide graphene nanoribbon (GNR) sections that are semimetallic, while the channel regions consist of narrow GNR sections that open semiconducting bandgaps. Our simulation employs a fully atomistic model of the device, contact and interfacial regions using tight-binding theory. The electronic structures are coupled with a self-consistent three-dimensional Poisson's equation to capture the nontrivial contact electrostatics, along with a quantum kinetic formulation of transport based on non-equilibrium Green's functions (NEGF). Although we only consider a specific device geometry, our results establish several general performance advantages of such monolithic devices (besides those related to fabrication and patterning), namely the improved electrostatics, suppressed short-channel effects, and Ohmic contacts at the narrow-to-wide interfaces.