Toward a Scalable Working Set Size Estimation Method and Its Application for Chip Multiprocessors

It is essential to accurately estimate the working set size (WSS) of an application for various optimizations such as to partition cache among virtual machines or reduce leakage power dissipated in an over-allocated cache by switching it OFF. However, the state-of-the-art heuristics such as average memory access latency (AMAL) or cache miss ratio (CMR) are poorly correlated to the WSS of an application due to 1) over-sized caches and 2) their dispersed nature. Past studies focus on estimating WSS of an application executing on a uniprocessor platform. Estimating the same for a chip multiprocessor (CMP) with a large dispersed cache is challenging due to the presence of concurrently executing threads/processes. Hence, we propose a scalable, highly accurate method to estimate WSS of an application. We call this method ``tagged WSS (TWSS)'' estimation method. We demonstrate the use of TWSS to switch-OFF the over-allocated cache ways in Static and Dynamic NonUniform Cache Architectures (SNUCA, DNUCA) on a tiled CMP. In our implementation of adaptable way SNUCA and DNUCA caches, decision of altering associativity is taken by each L2 controller. Hence, this approach scales better with the number of cores present on a CMP. It gives overall (geometric mean) 26% and 19% higher energy-delay product savings compared to AMAL and CMR heuristics on SNUCA, respectively.

Study of liquid fuel transport in a small carburetted engine in the context of cold-start HC emission control

In the present study, a detailed visualization of the transport of fuel film has been performed in a small carburetted engine with a transparent manifold at the exit of the carburettor. The presence of fuel film is observed significantly on the lower half of the manifold at idling, while at load conditions, the film is found to be distributed all throughout the manifold walls. Quantitative measurement of the fuel film in a specially-designed manifold of square cross section has also been performed using the planar laser-induced fluorescence (PLIF) technique. The measured fuel film thickness is observed to be of the order of 1 nun at idling, and in the range of 0.1 to 0.4 mm over the range of load and speed studied. These engine studies are complemented by experiments conducted in a carburettor rig to study the state of the fuel exiting the carburettor. Laser-based Particle/Droplet Image Analysis (PDIA) technique is used to identify fuel droplets and ligaments and estimate droplet diameters. At a throttle position corresponding to idling, the fuel exiting the carburettor is found to consist of very fine droplets of size less than 15 mu m and large fuel ligaments associated with length scales of the order of 500 mu m and higher. For a constant pressure difference across the carburettor, the fuel consists of droplets with an SMD of the order of 30 mu m. Also, the effect of liquid fuel film on the cold start HC emissions is studied. Based on the understanding obtained from these studies, strategies such as manifold heating and varying carburettor main jet nozzle diameter are implemented. These are observed to reduce emissions under both idling and varying load conditions.

Texture transition in cold-rolled nickel-40 wt.% cobalt alloy

The micromechanical aspects of rolling texture development in Ni-40 wt.% Co alloy during very large reductions (up to epsilon(t) = 3.9) have been studied. The alloy showed a typical Cu-type texture up to a true strain of epsilon(t) = 3; however, the texture undergoes an abrupt transition to Bs-type on further rolling to epsilon(t) approximate to 4. (The Bs-type texture, here, comprises almost equal fractions of Goss and Bs components.) Microstructural observations, at early stages, show that deformation is accommodated entirely by slip, and very little presence of deformation twinning is observed to explain the texture transition. However, at much higher reduction levels, micrographs show a high fraction of Cu-type shear bands. These bands are predominantly found in Cu-oriented grains and the crystallites inside the shear bands are preferentially oriented towards Goss, which could explain the final texture evolution. (C) 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Cold beam of isotopically pure Yb atoms by deflection using 1D-optical molasses

We demonstrate the generation of an isotopically pure beam of laser-cooled Yb atoms by deflection using 1D-optical molasses. Atoms in a collimated thermal beam are first slowed using a Zeeman slower. They are then subjected to a pair of molasses beams inclined at 45(a similar to) with respect to the slowed atomic beam. The slowed atoms are deflected and probed at a distance of 160 mm. We demonstrate the selective deflection of the bosonic isotope Yb-174 and the fermionic isotope Yb-171. Using a transient measurement after the molasses beams are turned on, we find a longitudinal temperature of 41 mK.

Design and Development of Thermoelectric Generator

In this paper we discuss the fabrication, working and characteristics of a thermoelectric generator made up of p and n type semiconductor materials. The device consists of Fe0.2Co3.8Sb11.5Te0.5 (zT = 1.04 at 818 K) as the n-type and Zn4Sb3 (zT= 0.8 at 550 K) as the p-type material synthesized by vacuum hot press method. Carbon paste has been used to join the semiconductor legs to metal (Molybdenum) electrodes to reduce the contact resistance. The multi-couple (4 legs) generator results a maximum output power of 1.083 mW at a temperature difference of 240 K between the hot and cold sides. In this investigation, an I-V characteristic, maximum output power of the thermoelectric module is presented. The efficiency of thermoelectric module is obtained as eta= 0.273 %.

Reply to ``Comment on `Strongly magnetized cold degenerate electron gas: Mass-radius relation of the magnetized white dwarf'''

We show that the upper bound for the central magnetic field of a super-Chandrasekhar white dwarf calculated by Nityananda and Konar Phys. Rev. D 89, 103017 (2014)] and in the concerned comment, by the same authors, against our work U. Das and B. Mukhopadhyay, Phys. Rev. D 86, 042001 (2012)] is erroneous. This in turn strengthens the argument in favor of the stability of the recently proposed magnetized super-Chandrasekhar white dwarfs. We also point out several other numerical errors in their work. Overall we conclude that the arguments put forth by Nityananda and Konar are misleading.

The cold mode: a phenomenological model for the evolution of density perturbations in the intracluster medium

Cool cluster cores are in global thermal equilibrium but are locally thermally unstable. We study a non-linear phenomenological model for the evolution of density perturbations in the intracluster medium (ICM) due to local thermal instability and gravity. We have analysed and extended a model for the evolution of an overdense blob in the ICM. We find two regimes in which the overdense blobs can cool to thermally stable low temperatures. One for large t(cool)/t(ff) (t(cool) is the cooling time and t(ff) is the free-fall time), where a large initial overdensity is required for thermal runaway to occur; this is the regime which was previously analysed in detail. We discover a second regime for t(cool)/t(ff) less than or similar to 1 (in agreement with Cartesian simulations of local thermal instability in an external gravitational field), where runaway cooling happens for arbitrarily small amplitudes. Numerical simulations have shown that cold gas condenses out more easily in a spherical geometry. We extend the analysis to include geometrical compression in weakly stratified atmospheres such as the ICM. With a single parameter, analogous to the mixing length, we are able to reproduce the results from numerical simulations; namely, small density perturbations lead to the condensation of extended cold filaments only if t(cool)/t(ff) less than or similar to 10.

Thermal Response on the Microstructure and Texture of ECAP and Cold-Rolled Pure Magnesium

This paper deals with dynamic recrystallization (DRX), static recrystallization, and grain growth phenomena of pure magnesium after equal channel angular pressing (ECAP) by route A and B-C at 523 K (250 A degrees C) followed by 80 pct cold rolling. The ECAP-deformed and the subsequently rolled samples were annealed at 373 K and 773 K (100 A degrees C and 500 A degrees C). The associated changes in the microstructure and texture were studied using electron back-scattered diffraction. ECAP produced an average grain size of 12 to 18 A mu m with B and C-2 fiber textures. Subsequent rolling led to an average grain size 8 to 10 A mu m with basal texture fiber parallel to ND. There was no noticeable increase in the average grain size on annealing at 373 K (100 A degrees C). However, significant increase in the average grain size occurred at 773 K (500 A degrees C). The occurrence of different DRX mechanisms was detected: discontinuous dynamic recrystallization was attributed to basal slip activity and continuous dynamic recovery and recrystallization to prismatic/pyramidal slip systems. Only continuous static recrystallization could be observed on annealing.

Continuous vapour adsorption cooling system with three adsorber beds

In this paper, the design of a new solar operated adsorption cooling system with two identical small and one large adsorber beds, which is capable of producing cold continuously, has been proposed. In this system, cold energy is stored in the form of refrigerant in a separate refrigerant storage tank at ambient temperature. Silica gel water is used as a working pair and system is driven by solar energy. The operating principle is described in details and its thermodynamic transient analysis is presented. Effect of COP and SCE for different adsorbent mass and adsorption/desorption time of smaller beds are discussed. Recommended mass and number of cycles of operation for smaller beds to attain continuous cooling with average COP and SCE of 0.63 and 337.5 kJ/kg, respectively are also discussed, at a generation, condenser and evaporator temperatures of 368 K, 303 K and 283 K, respectively. (C) 2015 Elsevier Ltd. All rights reserved.

COOL CORE CYCLES: COLD GAS AND AGN JET FEEDBACK IN CLUSTER CORES

Using high-resolution 3D and 2D (axisymmetric) hydrodynamic simulations in spherical geometry, we study the evolution of cool cluster cores heated by feedback-driven bipolar active galactic nuclei (AGNs) jets. Condensation of cold gas, and the consequent enhanced accretion, is required for AGN feedback to balance radiative cooling with reasonable efficiencies, and to match the observed cool core properties. A feedback efficiency (mechanical luminosity approximate to epsilon(M) over dot(acc)c(2); where (M) over dot(acc). is the mass accretion rate at 1 kpc) as small as 6 x 10(-5) is sufficient to reduce the cooling/accretion rate by similar to 10 compared to a pure cooling flow in clusters (with M-200 less than or similar to 7 x 10(14) M-circle dot). This value is much smaller compared to the ones considered earlier, and is consistent with the jet efficiency and the fact that only a small fraction of gas at 1 kpc is accreted onto the supermassive black hole (SMBH). The feedback efficiency in earlier works was so high that the cluster core reached equilibrium in a hot state without much precipitation, unlike what is observed in cool-core clusters. We find hysteresis cycles in all our simulations with cold mode feedback: condensation of cold gas when the ratio of the cooling-time to the free-fall time (t(cool)/t(ff)) is less than or similar to 10 leads to a sudden enhancement in the accretion rate; a large accretion rate causes strong jets and overheating of the hot intracluster medium such that t(cool)/t(ff) > 10; further condensation of cold gas is suppressed and the accretion rate falls, leading to slow cooling of the core and condensation of cold gas, restarting the cycle. Therefore, there is a spread in core properties, such as the jet power, accretion rate, for the same value of core entropy t(cool)/t(ff). A smaller number of cycles is observed for higher efficiencies and for lower mass halos because the core is overheated to a longer cooling time. The 3D simulations show the formation of a few-kpc scale, rotationally supported, massive (similar to 10(11) M-circle dot) cold gas torus. Since the torus gas is not accreted onto the SMBH, it is largely decoupled from the feedback cycle. The radially dominant cold gas (T < 5 x 10(4) K; vertical bar v(r)vertical bar >vertical bar v(phi vertical bar)) consists of fast cold gas uplifted by AGN jets and freely infalling cold gas condensing out of the core. The radially dominant cold gas extends out to 25 kpc for the fiducial run (halo mass 7 x 10(14) M-circle dot and feedback efficiency 6 x 10(-5)), with the average mass inflow rate dominating the outflow rate by a factor of approximate to 2. We compare our simulation results with recent observations.

Hybrid Working Set Algorithm for SVM Learning With a Kernel Coprocessor on FPGA

Support vector machines (SVM) are a popular class of supervised models in machine learning. The associated compute intensive learning algorithm limits their use in real-time applications. This paper presents a fully scalable architecture of a coprocessor, which can compute multiple rows of the kernel matrix in parallel. Further, we propose an extended variant of the popular decomposition technique, sequential minimal optimization, which we call hybrid working set (HWS) algorithm, to effectively utilize the benefits of cached kernel columns and the parallel computational power of the coprocessor. The coprocessor is implemented on Xilinx Virtex 7 field-programmable gate array-based VC707 board and achieves a speedup of upto 25x for kernel computation over single threaded computation on Intel Core i5. An application speedup of upto 15x over software implementation of LIBSVM and speedup of upto 23x over SVMLight is achieved using the HWS algorithm in unison with the coprocessor. The reduction in the number of iterations and sensitivity of the optimization time to variation in cache size using the HWS algorithm are also shown.

Cold gas in cluster cores: global stability analysis and non-linear simulations of thermal instability

We perform global linear stability analysis and idealized numerical simulations in global thermal balance to understand the condensation of cold gas from hot/virial atmospheres (coronae), in particular the intracluster medium (ICM). We pay particular attention to geometry (e.g. spherical versus plane-parallel) and the nature of the gravitational potential. Global linear analysis gives a similar value for the fastest growing thermal instability modes in spherical and Cartesian geometries. Simulations and observations suggest that cooling in haloes critically depends on the ratio of the cooling time to the free-fall time (t(cool)/t(ff)). Extended cold gas condenses out of the ICM only if this ratio is smaller than a threshold value close to 10. Previous works highlighted the difference between the nature of cold gas condensation in spherical and plane-parallel atmospheres; namely, cold gas condensation appeared easier in spherical atmospheres. This apparent difference due to geometry arises because the previous plane-parallel simulations focused on in situ condensation of multiphase gas but spherical simulations studied condensation anywhere in the box. Unlike previous claims, our non-linear simulations show that there are only minor differences in cold gas condensation, either in situ or anywhere, for different geometries. The amount of cold gas depends on the shape of tcool/tff; gas has more time to condense if gravitational acceleration decreases towards the centre. In our idealized plane-parallel simulations with heating balancing cooling in each layer, there can be significant mass/energy/momentum transfer across layers that can trigger condensation and drive tcool/tff far beyond the critical value close to 10.