Making STMs Cache Friendly with Compiler Transformations

Software transactional memory (STM) is a promising programming paradigm for shared memory multithreaded programs. In order for STMs to be adopted widely for performance critical software, understanding and improving the cache performance of applications running on STM becomes increasingly crucial, as the performance gap between processor and memory continues to grow. In this paper, we present the most detailed experimental evaluation to date, of the cache behavior of STM applications and quantify the impact of the different STM factors on the cache misses experienced by the applications. We find that STMs are not cache friendly, with the data cache stall cycles contributing to more than 50% of the execution cycles in a majority of the benchmarks. We find that on an average, misses occurring inside the STM account for 62% of total data cache miss latency cycles experienced by the applications and the cache performance is impacted adversely due to certain inherent characteristics of the STM itself. The above observations motivate us to propose a set of specific compiler transformations targeted at making the STMs cache friendly. We find that STM's fine grained and application unaware locking is a major contributor to its poor cache behavior. Hence we propose selective Lock Data co-location (LDC) and Redundant Lock Access Removal (RLAR) to address the lock access misses. We find that even transactions that are completely disjoint access parallel, suffer from costly coherence misses caused by the centralized global time stamp updates and hence we propose the Selective Per-Partition Time Stamp (SPTS) transformation to address this. We show that our transformations are effective in improving the cache behavior of STM applications by reducing the data cache miss latency by 20.15% to 37.14% and improving execution time by 18.32% to 33.12% in five of the 8 STAMP applications.

Solid State and Solution Mediated Multistep Sequential Transformations in Metal-Organic Coordination Networks

Sequential transformation in a family of metal-organic framework compounds has been investigated employing both a solid-state as well as a solution mediated route. The compounds, cobalt oxy-bis(benzoate) and manganese oxybis(benzoate) having a two-dimensional structure, were reacted with bipyridine forming cobalt oxy-bis(benzoate)-4,4'-bipyridine and manganese oxy-bis(benzoate)-4,4'-bipyridine, respectively. The bipyridine containing compounds appear to form sequentially through stable intermediates. For the cobalt system, the transformation from a two-dimensional compound, Co(H2O)(2)(OBA)] (OBA = 4,4'-oxy-bis(benzoate)), I, to two different three-dimensional compounds, Co(bpy)(OBA)]center dot bpy, II, (bpy = 4,4'-bipyridine) and Co(bpy)(0.5)(OBA)], III, and reversibility between II and III have been investigated. In the manganese system, transformation from a two-dimensional compound, Mn(H2O)(2)(OBA)], Ia, to two different three-dimensional compounds, Mn (bpy)(OBA)]center dot bpy, Ha and Ha to Mn(bpy)(0.5)(OBA)], Ilia, has been investigated. It has also been possible to identify intermediate products during these transformation reactions. The possible pathways for the formation of the compounds were postulated.

Loop transformations: convexity, pruning, and optimization

High-level loop transformations are a key instrument in mapping computational kernels to effectively exploit the resources in modern processor architectures. Nevertheless, selecting required compositions of loop transformations to achieve this remains a significantly challenging task; current compilers may be off by orders of magnitude in performance compared to hand-optimized programs. To address this fundamental challenge, we first present a convex characterization of all distinct, semantics-preserving, multidimensional affine transformations. We then bring together algebraic, algorithmic, and performance analysis results to design a tractable optimization algorithm over this highly expressive space. Our framework has been implemented and validated experimentally on a representative set of benchmarks running on state-of-the-art multi-core platforms.

Lead-free piezoelectric system (Na0.5Bi0.5)TiO3-BaTiO3: Equilibrium structures and irreversible structural transformations driven by electric field and mechanical impact

The structure-property correlation in the lead-free piezoelectric (1 - x)(Na0.5Bi0.5)TiO3-(x)BaTiO3 has been systematically investigated in detail as a function of composition (0 < x <= 0.11), temperature, electric field, and mechanical impact by Raman scattering, ferroelectric, piezoelectric measurement, x-ray, and neutron powder diffraction methods. Although x-ray diffraction study revealed three distinct composition ranges characterizing different structural features in the equilibrium state at room temperature: (i) monoclinic (Cc) + rhombohedral (R3c) for the precritical compositions, 0 <= x <= 0.05, (ii) cubiclike for 0.06 <= x <= 0.0675, and (iii) morphotropic phase boundary (MPB) like for 0.07 <= x < 0.10, Raman and neutron powder diffraction studies revealed identical symmetry for the cubiclike and the MPB compositions. The cubiclike structure undergoes irreversible phase separation by electric poling as well as by pure mechanical impact. This cubiclike phase exhibits relaxor ferroelectricity in its equilibrium state. The short coherence length (similar to 50A degrees) of the out-of-phase octahedral tilts does not allow the normal ferroelectric state to develop below the dipolar freezing temperature, forcing the system to remain in a dipolar glass state at room temperature. Electric poling helps the dipolar glass state to transform to a normal ferroelectric state with a concomitant enhancement in the correlation length of the out-of-phase octahedral tilt.

Component Selection in the Self-Assembly of Palladium(II) Nanocages and Cage-to-Cage Transformations

Dynamic supramolecular systems involving a tetratopic palladium(II) acceptor and three different pyridine-and imidazole-based donors have been used for self-selection by a synergistic effect of morphological information and coordination ability of ligands through specific coordination interactions. Three different cages were first synthesized by two-component self-assembly of individual donor and acceptor. When all four components were allowed to interact in a reaction mixture, only one out of three cages was isolated. The preferential binding affinity towards a particular partner was also established by transforming a non-preferred cage into a preferred cage by interaction with the appropriate ligand. Computational studies further supported the fact that coordination interaction of imidazole moiety to Pd-II is enthalpically more preferred compared to pyridine, which drives the selection process. Analysis of crystal packing of both complexes indicated the presence of strong hydrogen bonds between nitrate and water molecules and also H-bonded 3D networks of water. Both complexes exhibit promising proton conductivity (10(-5) to ca. 10(-3) Scm(-1)) at ambient temperature under a relative humidity of circa 98% with low activation energy.

Unraveling the nature of electric field- and stress- induced structural transformations in soft PZT by a new powder poling technique

A `powder-poling' technique was developed to study electric field induced structural transformations in ferroelectrics exhibiting a morphotropic phase boundary (MPB). The technique was employed on soft PZT exhibiting a large longitudinal piezoelectric response (d(33) similar to 650 pCN(-1)). It was found that electric poling brings about a considerable degree of irreversible tetragonal to monoclinic transformation. The same transformation was achieved after subjecting the specimen to mechanical stress, which suggests an equivalence of stress and electric field with regard to the structural mechanism in MPB compositions. The electric field induced structural transformation was also found to be accompanied by a decrease in the spatial coherence of polarization.

Structural and Magnetic Phase Transformations of Hydroxyapatite-Magnetite Composites under Inert and Ambient Sintering Atmospheres

The present work reports the impact of sintering conditions on the phase stability in hydroxyapatite (HA) magnetite (Fe3O4) bulk composites, which were densified using either pressureless sintering in air or by rapid densification via hot pressing in inert atmosphere. In particular, the phase abundances, structural and magnetic properties of the (1-x)HA-xFe(3)O(4) (x = 5, 10, 20, and 40 wt %) composites were quantified by corroborating results obtained from Rietveld refinement of the X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and Mossbauer spectroscopy. Post heat treatment phase analysis revealed a major retention of Fe3O4 in argon atmosphere, while it was partially/completely oxidized to hematite (alpha-Fe2O3) in air. Mossbauer results suggest the high-temperature diffusion of Fe3+ into hydroxyapatite lattice, leading to the formation of Fe-doped HA. A preferential occupancy of Fe3+ at the Ca(1) and Ca(2) sites under hot-pressing and conventional sintering conditions, respectively, was observed. The lattice expansion in HA from Rietveld analysis correlated well with the amounts of Fe-doped HA determined from the Mossbauer spectra. Furthermore, hydroxyapatite in the monoliths and composites was delineated to exist in the monoclinic (P2(1)/b) structure as against the widely reported hexagonal (P6(3)/m) crystal lattice. The compositional similarity of iron doping in hydroxyapatite to that of tooth enamel and bone presents HA-Fe3O4 composites as potential orthopedic and dental implant materials.

Composition dependent multiple structural transformations of myoglobin in aqueous ethanol solution: A combined experimental and theoretical study

Experimental studies (circular dichroism and ultra-violet (UV) absorption spectra) and large scale atomistic molecular dynamics simulations (accompanied by order parameter analyses) are combined to establish a number of remarkable (and unforeseen) structural transformations of protein myoglobin in aqueous ethanol mixture at various ethanol concentrations. The following results are particularly striking. (1) Two well-defined structural regimes, one at x(EtOH) similar to 0.05 and the other at x(EtOH) similar to 0.25, characterized by formation of distinct partially folded conformations and separated by a unique partially unfolded intermediate state at x(EtOH) similar to 0.15, are identified. (2) Existence of non-monotonic composition dependence of (i) radius of gyration, (ii) long range contact order, (iii) residue specific solvent accessible surface area of tryptophan, and (iv) circular dichroism spectra and UV-absorption peaks are observed. Interestingly at x(EtOH) similar to 0.15, time averaged value of the contact order parameter of the protein reaches a minimum, implying that this conformational state can be identified as a molten globule state. Multiple structural transformations well known in water-ethanol binary mixture appear to have considerably stronger effects on conformation and dynamics of the protein. We compare the present results with studies in water-dimethyl sulfoxide mixture where also distinct structural transformations are observed along with variation of co-solvent composition. (C) 2015 AIP Publishing LLC.

An alternative approach to study nonlinear inviscid flow over arbitrary bottom topography

This paper deals with a new approach to study the nonlinear inviscid flow over arbitrary bottom topography. The problem is formulated as a nonlinear boundary value problem which is reduced to a Dirichlet problem using certain transformations. The Dirichlet problem is solved by applying Plemelj-Sokhotski formulae and it is noticed that the solution of the Dirichlet problem depends on the solution of a coupled Fredholm integral equation of the second kind. These integral equations are solved numerically by using a modified method. The free-surface profile which is unknown at the outset is determined. Different kinds of bottom topographies are considered here to study the influence of bottom topography on the free-surface profile. The effects of the Froude number and the arbitrary bottom topography on the free-surface profile are demonstrated in graphical forms for the subcritical flow. Further, the nonlinear results are validated with the results available in the literature and compared with the results obtained by using linear theory. (C) 2015 Elsevier Inc. All rights reserved.

Observation of phase transformations in cement during hydration

We report the phase transformations in Portland cement before and after hydration. The hydration mechanism was studied in detail by using a full Rietveld refinement of the X-ray diffraction (XRD) patterns, Fourier Transformed Infra-Red (FTIR) spectroscopy, Thermogravimetric Analysis (TGA) and Mossbauer spectroscopy at room temperature. From the Rietveld refinement of XRD data, alite, belite, celite, brown-millerite and low quartz phases were detected and quantified as major phases in dry cement powder. After hydration, calcium carbonate, portlandite and ettringite phases were found to form. A large reduction in the amounts of alite and belite phases were observed suggesting the formation of amorphous C-S-H phase and emphasizing the role of alite phase in flash setting of cement, as justified by the XRD and FTIR spectroscopy. Mossbauer spectra of all the unset samples showed quadrupole split doublets corresponding to the brownmillerite phase which remains unchanged even after about one week of hydration, suggesting that brownmillerite did not transform to other phases during initial stage of hydration process. (C) 2015 Elsevier Ltd. All rights reserved.

PLUTO plus : Near-Complete Modeling of Affine Transformations for Parallelism and Locality

Affine transformations have proven to be very powerful for loop restructuring due to their ability to model a very wide range of transformations. A single multi-dimensional affine function can represent a long and complex sequence of simpler transformations. Existing affine transformation frameworks like the Pluto algorithm, that include a cost function for modern multicore architectures where coarse-grained parallelism and locality are crucial, consider only a sub-space of transformations to avoid a combinatorial explosion in finding the transformations. The ensuing practical tradeoffs lead to the exclusion of certain useful transformations, in particular, transformation compositions involving loop reversals and loop skewing by negative factors. In this paper, we propose an approach to address this limitation by modeling a much larger space of affine transformations in conjunction with the Pluto algorithm's cost function. We perform an experimental evaluation of both, the effect on compilation time, and performance of generated codes. The evaluation shows that our new framework, Pluto+, provides no degradation in performance in any of the Polybench benchmarks. For Lattice Boltzmann Method (LBM) codes with periodic boundary conditions, it provides a mean speedup of 1.33x over Pluto. We also show that Pluto+ does not increase compile times significantly. Experimental results on Polybench show that Pluto+ increases overall polyhedral source-to-source optimization time only by 15%. In cases where it improves execution time significantly, it increased polyhedral optimization time only by 2.04x.