On the Impact of Performance Faults in Modern Microprocessors

Modern microprocessors incorporate a variety of architectural features, such as branch prediction and speculative execution, which are not critical to the correctness of their operation yet are essential towards improving performance. Accordingly, while faults in the corresponding hardware may not necessarily affect functional correctness, they may, nevertheless, adversely impact performance. In this paper, we investigate quantitatively the performance impact of such faults using a superscalar, dynamically-scheduled, out-of-order, Alpha-like microprocessor, on which we execute SPEC2000 integer benchmarks. We provide extensive fault simulation-based experimental results that elucidate the various aspects of performance faults and we discuss how this information may guide the inclusion of additional hardware for performance loss recovery and yield enhancement.     

Characterization of Morphological Changes in Agglomerates Subject to Condensation and Evaporation Using Multiple Fractal Dimensions

Multiple fractal dimensions are used to characterize morphological changes that occur when an aerosol composed of irregularly shaped agglomerates is subject to condensation followed by evaporation. The agglomerates change from a branched, chainlike structure to a more regular, near-spherical or clumplike structure reflected in a decrease in the structural fractal dimension. The textural fractal dimension remains constant because the primary particles, of which the agglomerates are composed, do not change in shape. The degree of supersaturation and the number of condensation-evaporation cycles that the aerosol undergoes are major factors that influence morphological change. Even at low supersaturations, increasing the number of condensation-evaporation cycles makes the agglomerates more regular and thus decreases the structural fractal dimension. The transition point in the Richardson plot is a good indicator of the size of the primary particles in the agglomerate.     

Effect of channel geometry on solute dispersion in pressure-driven microfluidic systems

Pressure-driven transport of fluid and solute samples is often desirable in microfluidic devices, particularly where sufficient electroosmotic flow rates cannot be realized or the use of an electric field is restricted. Unfortunately, this mode of actuation also leads to hydrodynamic dispersion due to the inherent fluid shear in the system. While such dispersivity is known to scale with the square of the Peclet number based on the narrower dimension of the conduit (often the channel depth), the proportionality constant can vary significantly depending on its actual cross section. In this article, we review previous studies to understand the effect of commonly microfabricated channel cross sections on the Taylor–Aris dispersion of solute slugs in simple pressure-driven flow systems. We also analyze some recently proposed optimum designs which can reduce the contribution to this band broadening arising from the presence of the channel sidewalls. Finally, new simulation results have been presented in the last section of this paper which describe solutal spreading due to bowing of microchannels that can occur from stresses developed during their fabrication or operation under high-pressure conditions.     

UbiqStack: a taxonomy for a ubiquitous computing software stack

This paper describes a taxonomy for a ubiquitous computing software stack called UbiqStack. Through the lens of the UbiqStack taxonomy we survey a variety of subsystems designed to be the building blocks from which sophisticated infrastructures for ubiquitous computing are assembled. Our experience shows that many of these building blocks fit neatly into one of the five UbiqStack categories, each containing functionally-equivalent components. Effectively identifying the best-fit “Lego pieces”, which in turn determines the composite functionality of the resulting infrastructure, is critical. The selection process, however, is impeded by the lack of convention for labeling these classes of building blocks. The lack of clarity with respect to what ready-made subsystems are available within each class often results in naive re-implementation of ready-made components, monolithic and clumsy implementations, and implementations that impose non-standard interfaces onto the applications above. This paper describes the UbiqStack classes of subsystems and explores each in light of the experience gained over 2 years of active development of both ubiquitous computing applications and software infrastructures for their deployment.     

Rheological characterization of hydroxypropylcellulose gels

The present paper describes the rheological properties of hydroxypropylcellulose (HPC) gels formulated in propylene glycol (PG), water, ethanol, and mixtures of these components. The effects of molecular weight, polymer concentration, and solvent composition on the apparent viscosity and flow characteristics have been studied by continuous shear rheometry. The HPC gels are shear thinning and do not exhibit significant yield or hysteresis in their rheograms. The apparent viscosity increases with increasing molecular weight and concentration of the polymer, as expected. Although not so pronounced at lower concentrations (≤ 1.5%), HPC gels tend to become increasingly non-Newtonian with increasing molecular weight at higher polymer concentrations (3%). A mathematical model has been proposed for the prediction of viscosities of HPC gels. There exists a high degree of dependence on molecular interactions between various solvent molecules in the prediction of mixture viscosities in ternary systems. The effects of solvent composition on the viscoelastic behavior of these gels have also been examined by dynamic mechanical analysis. The HPC gels are highly viscoelastic and exhibit greater degrees of elasticity with increased PG content in ternary solvent mixtures with water and ethanol. The study also suggests that dynamic mechanical analysis could prove to be a useful tool in the determination of zero-shear viscosities, viscosities that are representative of most realistic situations.     

The specificity of the N-terminal SH2 domain of SHP-2 is modified by a single point mutation

SH2 domains are small protein domains of approximately 100 amino acids that bind to phosphotyrosine (pY) in the context of a specific sequence surrounding the target pY. In general, the residues C-terminal to the pY of the binding target are considered most important for defining the binding specificity, and in particular the pY + 1 and pY + 3 residues (i.e., the first and third amino acids C-terminal to the pY). However, our previous studies with the SH2 domains of the protein tyrosine phosphatase SHP-2 [Huyer, G., Li, Z. M., Adam, M., Huckle, W. R., and Ramachandran, C. (1995) Biochemistry 34, 1040-1049] indicated important interactions with the pY - 2 residue as well. In the SH2 domains of SHP-2, the highly conserved alphaA2 Arg is replaced by Gly. A comparison of the published crystal structures of the Src SH2 domain and the N-terminal SH2 domain of SHP-2 complexed with high-affinity peptides suggested that the alphaA2 Gly of SHP-2 creates a gap which is filled by the side chain of the pY - 2 residue of the bound peptide. It was predicted that replacing this Gly with Arg would alter or eliminate the involvement of the pY - 2 residue in binding. The alphaA2 Gly --> Arg mutant was constructed, and indeed, this mutant no longer required residues N-terminal to the target pY for high-affinity binding, making its specificity more like that of other SH2 domains. The alphaA2 Gly is clearly involved in directing the unusual requirement for the pY - 2 residue in the binding sequence of this SH2 domain, which has important implications for its in vivo targeting and specificity.     

Experiences with network-based user agents for mobile applications

Wireless networks are characterized by simple end devices and limited bandwidth. One solution to address these and other limitations of the wireless mobile environment that has been widely pursued is the placement of proxies, or agents, inside the network to assist with application processing that would normally take place on end devices. These agents can additionally manipulate data to reduce bandwidth requirements and assist in providing services. The design and implementation of a user agent is heavily impacted by the application it supports. In this paper we present our experiences with user agents in two different types of applications: telephony-based Personal Communication Services (PCS) and two-way messaging. To provide low latency service, the PCS user agents are mobile so that they may be kept close to their end devices. The design of the PCS agents limits migration overhead to the transfer of approximately 100 bytes of information in about 150 milliseconds. To provide advanced messaging services, the messaging user agent supports flexible messages while limiting air interface bandwidth usage. The messaging agents reduce uplink bandwidth usage by approximately 35% when applied to a cellular PCS system, and reduce message payloads by 97%. This revised version was published online in June 2006 with corrections to the Cover Date.     

Optimal Equalization of Discrete Signals Passed Through a Random Channel

The problem of equalizing a discrete signal that has been transmitted through a channel selected at random from an ensemble of channels is considered. Using mean-square error as the performance index, the minimum number of adjustable parameters required to achieve a given level of performance is sought. For certain special cases, it is shown that, using nonrecursive sampled data filters, the optimum tap weights are given by the eigenvectors of the matrix formed from the covariances of the channel's impulse response. A numerical algorithm is developed to find the optimum equalizer structure for a wide class of channels with the restriction that the number of channels in the given ensemble is finite. Results worked out for several examples show that the optimum equalizer structure requires significantly fewer adjustable parameters than the standard transversal equalizer in order to obtain the same level of performance.     

Laminar film boiling on a porous vertical wall with uniform suction velocity

Laminar free convection film boiling on a porous vertical wall with uniform suction or blowing is analysed using boundary layer theory. The solutions are obtained assuming suction or blowing to be a disturbance superposed on the isothermal, impermeable wall case. Using a parameter involving the suction or blowing velocity, universal functions are derived for various values of Prandtl Number and c_p(T_w — T_sat)/h_fgPr. These universal functions can be used to estimate the heat transfer rate in the presence of suction or blowing. As expected, suction increases the heat transfer rate while blowing decreases the heat-transfer. Even small velocities of suction or blowing could significantly affect the heat transfer. It is also found that the effects of suction or blowing are more pronounced at lower wall superheats.     

Significance of low water—solid ratio and temperature on the physico-mechanical characteristics of hydrates of tricalcium aluminate

Hydration characteristics are reported of tricalcium aluminate studied at 20 and 80 °C using water-solid ratios of 0.2 and 1.0. Hydration products were subjected to differential thermal analysis, thermogravimetric analysis, scanning electron microscopy, length change, porosity and pore size distribution, and micro-hardness measurements. Hydration proceeds faster at 80 °C than at 20 °C and the cubic C₃AH₆ phase is detected in the first few seconds. At 20 °C hydration occurs with the initial formation of the hexagonal phases and expansion is continuous. At higher temperatures and a water-solid ratio of 0.2, after an initial small expansion, the dimensional change is low. The product at 80 °C develops more than fourfold the hardness value developed at 20 °C. After 2 days of hydration, porosity, per cent C₃AH₆ and microhardness are 15.2%, 83% and 38.9 kg/mm²; corresponding values for the product obtained at 20 °C are 21.5%, 75% and 9.1 kg/mm². Microstructural examination of the material formed at 80 °C indicates a closely welded and continuous network of the cubic phase. These results reveal that at a low water-solid ratio and higher temperatures the formation of the cubic phase from C₃A results in an enhancement of strength.