Compiler-Directed Frequency and Voltage Scaling for a Multiple Clock Domain

Multiple Clock Domain processors provide an attractive solution to the increasingly challenging problems of clock distribution and power dissipation. They allow their chips to be partitioned into different clock domains, and each domain’s frequency (voltage) to be independently configured. This flexibility adds new dimensions to the Dynamic Voltage and Frequency Scaling problem, while providing better scope for saving energy and meeting performance demands. In this paper, we propose a compiler directed approach for MCD-DVFS. We build a formal petri net based program performance model, parameterized by settings of microarchitectural components and resource configurations, and integrate it with our compiler passes for frequency selection.Our model estimates the performance impact of a frequency setting, unlike the existing best techniques which rely on weaker indicators of domain performance such as queue occupancies(used by online methods) and slack manifestation for a particular frequency setting (software based methods).We evaluate our method with subsets of SPECFP2000,Mediabench and Mibench benchmarks. Our mean energy savings is 60.39% (versus 33.91% of the best software technique)in a memory constrained system for cache miss dominated benchmarks, and we meet the performance demands.Our ED2 improves by 22.11% (versus 18.34%) for other benchmarks. For a CPU with restricted frequency settings, our energy consumption is within 4.69% of the optimal.

Compiler-Directed Dynamic Voltage Scaling using Program Phases

Energy consumption has become a major constraint in providing increased functionality for devices with small form factors. Dynamic voltage and frequency scaling has been identified as an effective approach for reducing the energy consumption of embedded systems. Earlier works on dynamic voltage scaling focused mainly on performing voltage scaling when the CPU is waiting for memory subsystem or concentrated chiefly on loop nests and/or subroutine calls having sufficient number of dynamic instructions. This paper concentrates on coarser program regions and for the first time uses program phase behavior for performing dynamic voltage scaling. Program phases are annotated at compile time with mode switch instructions. Further, we relate the Dynamic Voltage Scaling Problem to the Multiple Choice Knapsack Problem, and use well known heuristics to solve it efficiently. Also, we develop a simple integer linear program formulation for this problem. Experimental evaluation on a set of media applications reveal that our heuristic method obtains a 38% reduction in energy consumption on an average, with a performance degradation of 1% and upto 45% reduction in energy with a performance degradation of 5%. Further, the energy consumed by the heuristic solution is within 1% of the optimal solution obtained from the ILP approach.

Universality of scaling laws in correlation between velocity and shear stress in turbulent boundary layers

In this study, we analyse simultaneous measurements (at 50 Hz) of velocity at several heights and shear stress at the surface made during the Utah field campaign for the presence of ranges of scales, where distinct scale-to-scale interactions between velocity and shear stress can be identified. We find that our results are similar to those obtained in a previous study [Venugopal et al., 2003] (contrary to the claim in V2003, that the scaling relations might be dependent on Reynolds number) where wind tunnel measurements of velocity and shear stress were analysed. We use a wavelet-based scale-to-scale cross-correlation to detect three ranges of scales of interaction between velocity and shear stress, namely, (a) inertial subrange, where the correlation is negligible; (b) energy production range, where the correlation follows a logarithmic law; and (c) for scales larger than the boundary layer height, the correlation reaches a plateau.

Dynamic Supply and Threshold Voltage Scaling for CMOS Digital Circuits Using In-Situ Power Monitor

A generalized power tracking algorithm that minimizes power consumption of digital circuits by dynamic control of supply voltage and the body bias is proposed. A direct power monitoring scheme is proposed that does not need any replica and hence can sense total power consumed by load circuit across process, voltage, and temperature corners. Design details and performance of power monitor and tracking algorithm are examined by a simulation framework developed using UMC 90-nm CMOS triple well process. The proposed algorithm with direct power monitor achieves a power savings of 42.2% for activity of 0.02 and 22.4% for activity of 0.04. Experimental results from test chip fabricated in AMS 350 nm process shows power savings of 46.3% and 65% for load circuit operating in super threshold and near sub-threshold region, respectively. Measured resolution of power monitor is around 0.25 mV and it has a power overhead of 2.2% of die power. Issues with loop convergence and design tradeoff for power monitor are also discussed in this paper.

Appraisal of Bishop's method of slope stability analysis

The stability of slopes is a major problem in geotechnical engineering. Of the methods available for the analysis of soil slopes such as limit equilibrium methods, limit analysis and numerical methods such as FEM and FDM, limit equilibrium methods are popular and generally used, owing to their simplicity in formulation and in evaluating the overall factor of safety of slope. However limit equilibrium methods possess certain disadvantages. They do not consider whether the slope is an embankment or natural slope or an excavation and ignore the effect of incremental construction, initial stress, stress strain behavior etc. In the work reported in this paper, a comparative study of actual state of stress and actual factor of safety and Bishop's factor of safety is performed. The actual factor of safety is obtained by consideration of contours of mobilised shear strains. Using Bishop's method of slices, the critical slip surfaces of a number of soil slopes with different geometries are determined and both the factors of safety are obtained. The actual normal stresses and shear stresses are determined from finite difference formulation using FLAG (Fast Lagrangian Analysis of Continuaa) with Mohr-Coulomb model. The comparative study is performed in terms of parameter lambda(c phi) (= gamma H tan phi/c). I is shown that actual factor of safety is higher than Bishop's factor of safety depending on slope angle and lambda(c phi).

Effect of strength - Matric suction variations on stability of a natural slope

Stability analysis of residual soil slopes are now increasingly being performed with the incorporation of the matric suction component of strength. The matric suction (u(a)-u(w)) component of shear strength is known as apparent cohesion. The relation between matric suction and apparent cohesion (c(app)) may be linear or non-linear. The impact of type of apparent strength versus matric suction relationship on the stability of an unsaturated residual soil slope is examined in this study. Results of the study showed that the factor of safety values were unaffected by the nature of the strength versus matric suction relationship for the residual soil slope examined. This was so as contribution from the effective stress- strength component to the factor of safety predominated over the contribution made by the apparent strength component.

Effect of scaling on the non-quasi-static behaviour of the MOSFET for RF ICs

The effect of scaling (1 μm to 0.09 μm) on the non-quasi-static (NQS) behaviour of the MOSFET has been studied using process and device simulation. It is shown that under fixed gate (Vgs) and drain (Vds) bias voltages, the NQS transition frequency (fNQS) scales as 1/Leff rather than 1/L2eff due to the velocity saturation effect. However, under the practical scaling guidelines, considering the scaling of supply voltage as well, fNQS shows a turn around effect at the sub 100 nm regime. The relation between unity gain frequency (ft) and fNQS is also evaluated and it is shown that ft and fNQS have similar trends with scaling.

Dynamic Supply and Threshold Voltage Scaling for CMOS Digital Circuits Using In-Situ Power Monitor

A generalized power tracking algorithm that minimizes power consumption of digital circuits by dynamic control of supply voltage and the body bias is proposed. A direct power monitoring scheme is proposed that does not need any replica and hence can sense total power consumed by load circuit across process, voltage, and temperature corners. Design details and performance of power monitor and tracking algorithm are examined by a simulation framework developed using UMC 90-nm CMOS triple well process. The proposed algorithm with direct power monitor achieves a power savings of 42.2% for activity of 0.02 and 22.4% for activity of 0.04. Experimental results from test chip fabricated in AMS 350 nm process shows power savings of 46.3% and 65% for load circuit operating in super threshold and near sub-threshold region, respectively. Measured resolution of power monitor is around 0.25 mV and it has a power overhead of 2.2% of die power. Issues with loop convergence and design tradeoff for power monitor are also discussed in this paper.

Quenching across quantum critical points in periodic systems: Dependence of scaling laws on periodicity

We study the quenching dynamics of a many-body system in one dimension described by a Hamiltonian that has spatial periodicity. Specifically, we consider a spin-1/2 chain with equal xx and yy couplings and subject to a periodically varying magnetic field in the (z) over cap direction or, equivalently, a tight-binding model of spinless fermions with a periodic local chemical potential, having period 2q, where q is a positive integer. For a linear quench of the strength of the magnetic field (or chemical potential) at a rate 1/tau across a quantum critical point, we find that the density of defects thereby produced scales as 1/tau(q/(q+1)), deviating from the 1/root tau scaling that is ubiquitous in a range of systems. We analyze this behavior by mapping the low-energy physics of the system to a set of fermionic two-level systems labeled by the lattice momentum k undergoing a nonlinear quench as well as by performing numerical simulations. We also show that if the magnetic field is a superposition of different periods, the power law depends only on the smallest period for very large values of tau, although it may exhibit a crossover at intermediate values of tau. Finally, for the case where a zz coupling is also present in the spin chain, or equivalently, where interactions are present in the fermionic system, we argue that the power associated with the scaling law depends on a combination of q and the interaction strength.

Fatigue crack growth due to overloads in plain concrete using scaling laws

Scaling laws are represented in power law form and can be utilized to extract the characteristic properties of a new phenomenon with the help of self-similar solutions. In this work, an attempt has been made to propose a scaling law analytically, for plain concrete when subjected to variable amplitude loading. Due to the application of overload on concrete structures, acceleration in the crack growth process takes place. A closed form expression has been developed to capture the acceleration in crack growth rate in conjunction with the principles of dimensional analysis and self-similarity. The proposed model accounts for parameters such as, the tensile strength, fracture toughness, overload effect and the structural size. Knowing the governed and the governing parameters of the physical problem and by using the concepts of self-similarity, a relationship is obtained between the different parameters involved. The predicted results are compared with experimental crack growth data for variable amplitude loading and are found to capture the overload effect with sufficient accuracy. Through a sensitivity analysis, fracture toughness is found to be the most dominant parameter in accelerating the crack length due to application of overload.

Scaling analysis: Equivalence of convective and radiative heating of levitated droplet

This letter develops theoretical relationships for equilibrium timescale and temperature scale of a vaporizing droplet in a convective and a radiative environment. The transient temperature normalized by the respective scales exhibits a unified profile for both modes of heating. The analysis allows for the prediction of the required laser flux to show its equivalence in a corresponding heated gas stream. The theoretical equivalence shows good agreement with experiments across a range of droplet sizes. Simple experiments can be conducted in a levitator to extrapolate information in realistic convective environments like combustion and spray drying. (C) 2012 American Institute of Physics. http://dx.doi.org/10.1063/1.4720092]

Relative effect of slope and equilibrium line altitude on the retreat of Himalayan glaciers

It has been observed that a majority of glaciers in the Himalayas have been retreating. In this paper, we show that there are two major factors which control the advance/retreat of the Himalayan glaciers. They are the slope of the glacier and changes in the equilibrium line altitude. While it is well known, that these factors are important, we propose a new way of combining them and use it to predict retreat. The functional form of this model has been derived from numerical simulations using an ice-flow code. The model has been successfully applied to the movement of eight Himalayan glaciers during the past 25 years. It explains why the Gangotri glacier is retreating while Zemu of nearly the same length is stationary, even if they are subject to similar environmental changes. The model has also been applied to a larger set of glaciers in the Parbati basin, for which retreat based on satellite data is available, though over a shorter time period.

Microstructural Evolution of A356 Al Alloy During Flow Along a Cooling Slope

Cooling slope (CS) has been used in this study to prepare semi-solid slurry of A356 Al alloy, keeping in view of slurry generation on demand for Rheo-pressure die casting process. Understanding the physics of microstructure evolution during cooling slope slurry formation is important to satisfy the need of semi-sold slurry with desired shape, size and morphology of primary Al phase. Mixture of spherical and rosette shaped primary Al phase has been observed in the samples collected during melt flow through the slope as well as in the cast (mould) samples compared to that of dendritic shape, observed in case of conventionally cast A356 alloy. The liquid melt has been poured into the slope at 650 A degrees C temperature and during flow it falls below the liquidus temperature of the said alloy, which facilitates crystallization of alpha-Al crystals on the cooling slope wall. Crystal separation due to melt flow is found responsible for nearly spherical morphology of the primary Al phase.