Algorithmic skeleton framework for the orchestration of GPU computations

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

The Loop-of-Stencil-Reduce Paradigm

In this paper we advocate the Loop-of-stencil-reduce pattern as a way to simplify the parallel programming of heterogeneous platforms (multicore+GPUs). Loop-of-Stencil-reduce is general enough to subsume map, reduce, map-reduce, stencil, stencil-reduce, and, crucially, their usage in a loop. It transparently targets (by using OpenCL) combinations of CPU cores and GPUs, and it makes it possible to simplify the deployment of a single stencil computation kernel on different GPUs. The paper discusses the implementation of Loop-of-stencil-reduce within the FastFlow parallel framework, considering a simple iterative data-parallel application as running example (Game of Life) and a highly effective parallel filter for visual data restoration to assess performance. Thanks to the high-level design of the Loop-of-stencil-reduce, it was possible to run the filter seamlessly on a multicore machine, on multi-GPUs, and on both.

Multi-GPU support on the marrow algorithmic skeleton framework

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

Heterogeneous computing with an algorithmic skeleton framework

The Graphics Processing Unit (GPU) is present in almost every modern day personal computer. Despite its specific purpose design, they have been increasingly used for general computations with very good results. Hence, there is a growing effort from the community to seamlessly integrate this kind of devices in everyday computing. However, to fully exploit the potential of a system comprising GPUs and CPUs, these devices should be presented to the programmer as a single platform. The efficient combination of the power of CPU and GPU devices is highly dependent on each device’s characteristics, resulting in platform specific applications that cannot be ported to different systems. Also, the most efficient work balance among devices is highly dependable on the computations to be performed and respective data sizes. In this work, we propose a solution for heterogeneous environments based on the abstraction level provided by algorithmic skeletons. Our goal is to take full advantage of the power of all CPU and GPU devices present in a system, without the need for different kernel implementations nor explicit work-distribution.To that end, we extended Marrow, an algorithmic skeleton framework for multi-GPUs, to support CPU computations and efficiently balance the work-load between devices. Our approach is based on an offline training execution that identifies the ideal work balance and platform configurations for a given application and input data size. The evaluation of this work shows that the combination of CPU and GPU devices can significantly boost the performance of our benchmarks in the tested environments, when compared to GPU-only executions.

Towards an algorithmic skeleton framework for programming the Intel R Xeon PhiTM processor

The Intel R Xeon PhiTM is the first processor based on Intel’s MIC (Many Integrated Cores) architecture. It is a co-processor specially tailored for data-parallel computations, whose basic architectural design is similar to the ones of GPUs (Graphics Processing Units), leveraging the use of many integrated low computational cores to perform parallel computations. The main novelty of the MIC architecture, relatively to GPUs, is its compatibility with the Intel x86 architecture. This enables the use of many of the tools commonly available for the parallel programming of x86-based architectures, which may lead to a smaller learning curve. However, programming the Xeon Phi still entails aspects intrinsic to accelerator-based computing, in general, and to the MIC architecture, in particular. In this thesis we advocate the use of algorithmic skeletons for programming the Xeon Phi. Algorithmic skeletons abstract the complexity inherent to parallel programming, hiding details such as resource management, parallel decomposition, inter-execution flow communication, thus removing these concerns from the programmer’s mind. In this context, the goal of the thesis is to lay the foundations for the development of a simple but powerful and efficient skeleton framework for the programming of the Xeon Phi processor. For this purpose we build upon Marrow, an existing framework for the orchestration of OpenCLTM computations in multi-GPU and CPU environments. We extend Marrow to execute both OpenCL and C++ parallel computations on the Xeon Phi. We evaluate the newly developed framework, several well-known benchmarks, like Saxpy and N-Body, will be used to compare, not only its performance to the existing framework when executing on the co-processor, but also to assess the performance on the Xeon Phi versus a multi-GPU environment.

A rapidly evolving secretome builds and patterns a sea shell

Background: Instructions to fabricate mineralized structures with distinct nanoscale architectures, such as seashells and coral and vertebrate skeletons, are encoded in the genomes of a wide variety of animals. In mollusks, the mantle is responsible for the extracellular production of the shell, directing the ordered biomineralization of CaCO3 and the deposition of architectural and color patterns. The evolutionary origins of the ability to synthesize calcified structures across various metazoan taxa remain obscure, with only a small number of protein families identified from molluskan shells. The recent sequencing of a wide range of metazoan genomes coupled with the analysis of gene expression in non-model animals has allowed us to investigate the evolution and process of biomineralization in gastropod mollusks. Results: Here we show that over 25% of the genes expressed in the mantle of the vetigastropod Haliotis asinina encode secreted proteins, indicating that hundreds of proteins are likely to be contributing to shell fabrication and patterning. Almost 85% of the secretome encodes novel proteins; remarkably, only 19% of these have identifiable homologues in the full genome of the patellogastropod Lottia scutum. The spatial expression profiles of mantle genes that belong to the secretome is restricted to discrete mantle zones, with each zone responsible for the fabrication of one of the structural layers of the shell. Patterned expression of a subset of genes along the length of the mantle is indicative of roles in shell ornamentation. For example, Has-sometsuke maps precisely to pigmentation patterns in the shell, providing the first case of a gene product to be involved in molluskan shell pigmentation. We also describe the expression of two novel genes involved in nacre (mother of pearl) deposition. Conclusion: The unexpected complexity and evolvability of this secretome and the modular design of the molluskan mantle enables diversification of shell strength and design, and as such must contribute to the variety of adaptive architectures and colors found in mollusk shells. The composition of this novel mantle-specific secretome suggests that there are significant molecular differences in the ways in which gastropods synthesize their shells.

(Table T1) Abundance of radiolarian skeletons in scrape samples of ODP Hole 178-1098B

Palmer Deep is a series of three glacially overdeepened basins on the Antarctic Peninsula shelf, ~20 km southwest of Anvers Island. Site 1098 (64°51.72'S, 64°12.48'W) was drilled in the shallowest basin, Basin I, at 1012 m water depth. The sediment recovered was primarily laminated, siliceous, biogenic, pelagic muds alternating with siliciclastic hemipelagic sediments (Barker, Camerlenghi, Acton, et al., 1999). Sedimentation rates of 0.1725 cm/yr in the upper 25 m and 0.7-0.80 cm/yr in the lower 25 m of the core have been estimated from 14C (Domack et al., 2001). The oldest datable sediments have an age of ~13 ka and were underlain by diamicton sediments of the last glacial maximum (Domack et al., 2001). The large-scale water-mass distribution and circulation in the vicinity of Palmer Deep is dominated by Circumpolar Deep Water (CDW) below 200 m (Hofmann et al., 1996). Palmer Deep is too far from the coast to be influenced by glacial meltwater and cold-tongue generation associated with it (Domack and Williams, 1990; Dixon and Domack, 1991). Circulation patterns in the Palmer Deep area are not well understood, but evidence suggests southward flow across Palmer Deep from Anvers Island to Renaud Island (Kock and Stein, 1978). The water south of Anvers Island is nearly open with loose pack ice from February through May. The area is covered with sea ice beginning in June (Gloersen et al., 1992; Leventer et al., 1996). Micropaleontologic data from the work of Leventer et al. (1996) on a 9-m piston core has revealed circulation and climate patterns for the past 3700 yr in the Palmer Deep. The benthic foraminifer assemblage is dominated by two taxa, Bulimina aculeata and Bolivina pseudopunctata, which are inversely related. High relative abundances of B. aculeata occur cyclically over a period of ~230 yr. The assemblage associated with high abundance of B. aculeata in Palmer Deep resembles that from the Bellingshausen shelf, which is associated with CDW. In addition to the faunal evidence, hydrographic data indicate incursions of CDW into Palmer Deep (Leventer et al., 1996). A distinctive diatom assemblage dominated by a single genus was associated with peaks in B. aculeata, whereas a few different assemblages were associated with lows in B. aculeata. Leventer et al. (1996) interpreted the variability in diatom assemblages as an indication of changes in productivity associated with changes in water column stability. Abelmann and Gowing (1997) studied the horizontal and vertical distributions of radiolarians in the Atlantic sector of the Southern Ocean. They show that the spatial distribution of radiolarian assemblages reflects hydrographic boundaries. In a transect from the subtropical Atlantic to polar Antarctic zones, radiolarians in the upper 1000 m of the water column occurred in distinct surface and deep-living assemblages related to water depth, temperature, salinity, and nutrient content. Living assemblages resembled those preserved in underlying surface sediments (Abelmann and Gowing, 1997). Circumantarctic coastal sediments from neritic environments contained a distinctive assemblage dominated by the Phormacantha hystrix/Plectacantha oikiskos group and Rhizoplegma boreale (Nishimura et al., 1997). Low diversity and species compositions distinguished the coastal sediments from the typical pelagic Antarctic assemblages. Factors that controlled the assemblages were water depth, proximity to the coast, occurrence of sea ice, and steepness of topography, rather than temperature and salinity. Nishimura et al. (1997) found a gradient of sorts from deep-water sites containing diverse assemblages typical of pelagic environments to coastal sites with low diversity assemblages dominated by P. hystrix/P. oikiskos group and R. boreale. In general, sites between these two extremes had increased proportions of the coastal assemblage with decreasing water depth (Nishimura et al., 1997). At a site near Hole 1098 (GC905), they showed that the relative abundance of the coastal assemblage increased downcore (Nishimura et al., 1997). The purpose of the research presented here was to make a cursory investigation into the radiolarian assemblages as possible paleoenvironmental indicators.

A green perspective on Structured parallel programming

Structured parallel programming, and in particular programming models using the algorithmic skeleton or parallel design pattern concepts, are increasingly considered to be the only viable means of supporting effective development of scalable and efficient parallel programs. Structured parallel programming models have been assessed in a number of works in the context of performance. In this paper we consider how the use of structured parallel programming models allows knowledge of the parallel patterns present to be harnessed to address both performance and energy consumption. We consider different features of structured parallel programming that may be leveraged to impact the performance/energy trade-off and we discuss a preliminary set of experiments validating our claims.

Characterization Of Cumulative Joint Damage Patterns In Patients With Rheumatoid Arthritis: A Clinical, Serological, And Gene Polymorphism Perspective.

To characterize cumulative joint damage (CJD) patterns in rheumatoid arthritis (RA) and determine their associations with demographic/clinical features and HLA-DRB1 gene polymorphism. Hand and foot radiographs were obtained from 404 patients with RA. CJD patterns were determined by 3 derivations from Sharp/van der Heijde scores, obtained by the mathematical division of scores for hands/feet (Sharp-h/f score), fingers/wrists (Sharp-f/w score), and erosion/space narrowing (Sharp-e/sn score), respectively. DNA and serum were obtained for determination of HLA-DRB1 polymorphism, rheumatoid factor (RF), and anticitrullinated protein antibodies (ACPA). Patients with wrist-dominant CJD pattern were more likely to have severe RA than those with finger-dominant pattern (68.4% vs 46.0%; p = 0.036) as were those with foot-dominant vs hand-dominant CJD pattern (76.5% vs 56.4%; p = 0.044). HLA-DRB1 shared epitope (SE) alleles were associated with erosion-dominant CJD pattern (p = 0.021). Patients with erosion-dominant CJD pattern had higher levels of RF and ACPA than those with space-narrowing-dominant CJD pattern (median RF 71.35 U/ml vs 22.05 U/ml, respectively; p = 0.003; median ACPA 187.9 U/ml vs 143.2 U/ml, respectively; p < 0.001). The majority of triple-positive patients (SE+, RF+, ACPA+) had erosion-dominant CJD pattern (62.3%) while the majority of triple-negative patients (SE-, FR-, ACPA-) had space narrowing-dominant CJD pattern (75%; p = 0.017). ACPA was associated with HLA-DRB1 SE alleles (p < 0.05). Patients with foot-dominant CJD pattern were taller than those with hand-dominant CJD pattern (p = 0.002); those with erosion-dominant CJD pattern had higher weight and body mass index than those with space narrowing-dominant CJD pattern (p = 0.014, p = 0.001). CJD patterns were associated with disease severity, HLA-DRB1 SE status, presence and titer of ACPA and RF, and morphometric features.

Environmental Controls In The Water Use Patterns Of A Tropical Cloud Forest Tree Species, Drimys Brasiliensis (winteraceae).

Trees from tropical montane cloud forest (TMCF) display very dynamic patterns of water use. They are capable of downwards water transport towards the soil during leaf-wetting events, likely a consequence of foliar water uptake (FWU), as well as high rates of night-time transpiration (Enight) during drier nights. These two processes might represent important sources of water losses and gains to the plant, but little is known about the environmental factors controlling these water fluxes. We evaluated how contrasting atmospheric and soil water conditions control diurnal, nocturnal and seasonal dynamics of sap flow in Drimys brasiliensis (Miers), a common Neotropical cloud forest species. We monitored the seasonal variation of soil water content, micrometeorological conditions and sap flow of D. brasiliensis trees in the field during wet and dry seasons. We also conducted a greenhouse experiment exposing D. brasiliensis saplings under contrasting soil water conditions to deuterium-labelled fog water. We found that during the night D. brasiliensis possesses heightened stomatal sensitivity to soil drought and vapour pressure deficit, which reduces night-time water loss. Leaf-wetting events had a strong suppressive effect on tree transpiration (E). Foliar water uptake increased in magnitude with drier soil and during longer leaf-wetting events. The difference between diurnal and nocturnal stomatal behaviour in D. brasiliensis could be attributed to an optimization of carbon gain when leaves are dry, as well as minimization of nocturnal water loss. The leaf-wetting events on the other hand seem important to D. brasiliensis water balance, especially during soil droughts, both by suppressing tree transpiration (E) and as a small additional water supply through FWU. Our results suggest that decreases in leaf-wetting events in TMCF might increase D. brasiliensis water loss and decrease its water gains, which could compromise its ecophysiological performance and survival during dry periods.

Patterns Of Seizure Control In Patients With Mesial Temporal Lobe Epilepsy With And Without Hippocampus Sclerosis.

Objective Patients with mesial temporal lobe epilepsy (MTLE) may present unstable pattern of seizures. We aimed to evaluate the occurrence of relapse-remitting seizures in MTLE with (MTLE-HS) and without (MTLE-NL) hippocampal sclerosis. Method We evaluated 172 patients with MTLE-HS (122) or MTLE-NL (50). Relapse-remitting pattern was defined as periods longer than two years of seizure-freedom intercalated with seizure recurrence. Infrequent seizures was considered as up to three seizures per year and frequent seizures as any period of seizures higher than that. Results Thirty-seven (30%) MTLE-HS and 18 (36%) MTLE-NL patients had relapse-remitting pattern (X2, p = 0.470). This was more common in those with infrequent seizures (X2, p < 0.001). Twelve MTLE-HS and one MTLE-NL patients had prolonged seizure remission between the first and second decade of life (X2, p = 0.06). Conclusion Similar proportion of MTLE-HS or MTLE-NL patients present relapse-remitting seizures and this occurs more often in those with infrequent seizures.

Breastfeeding and non-nutritive sucking patterns related to the prevalence of anterior open bite in primary dentition

OBJECTIVE: Nutritional, immunological and psychological benefts of exclusive breastfeeding for the frst 6 months of life are unequivocally recognized. However, mothers should also be aware of the importance of breastfeeding for promoting adequate oral development. This study evaluated the association between breastfeeding and non-nutritive sucking patterns and the prevalence of anterior open bite in primary dentition. MATERIAL AND METHODS: Infant feeding and non-nutritive sucking were investigated in a 3-6 year-old sample of 1,377 children, from São Paulo city, Brazil. Children were grouped according to breastfeeding duration: G1 - non-breastfed, G2 - shorter than 6 months, G3 - interruption between 6 and 12 months, and G4 - longer than 12 months. Three calibrated dentists performed clinical examinations and classifed overbite into 3 categories: normal, anterior open bite and deep bite. Chi-square tests (p<0.05) with odds ratio (OR) calculation were used for intergroup comparisons. The impact of breastfeeding and non-nutritive sucking on the prevalence of anterior open bite was analyzed using binary logistic regression. RESULTS: The prevalence estimates of anterior open bite were: 31.9% (G1), 26.1% (G2), 22.1% (G3), and 6.2% (G4). G1 would have signifcantly more chances of having anterior open bite compared with G4; in the total sample (OR=7.1) and in the subgroup without history of non-nutritive sucking (OR=9.3). Prolonging breastfeeding for 12 months was associated with a 3.7 times lower chance of having anterior open bite. In each year of persistence with non-nutritive sucking habits, the chance of developing this malocclusion increased in 2.38 times. CONCLUSIONS: Breastfeeding and non-nutritive sucking durations demonstrated opposite effects on the prediction of anterior open bite. Non-breastfed children presented signifcantly greater chances of having anterior open bite compared with those who were breastfed for periods longer than 12 months, demonstrating the benefcial infuence of breastfeeding on dental occlusion.

Polyamine patterns in haploid and diploid tobacco tissues and in vitro cultures

The aim of this work was to determine PAs levels in pith tissues and callus cultures from haploid and diploid tobacco plants, explanted from the apical and basal regions of the stem. These explants were cultured in an RM-64 medium supplied with IAA and kinetin, under light or in the dark, during successive subcultures. PAs levels followed a basipetal decrease in diploid and an increase in haploid, pith tissues. A similar pattern of total PAs (free + conjugated) was observed for the callus of diploid and haploid plants maintained in the light, and for the haploid callus in the dark, whereas the diploid callus in the dark showed a constant increase in total PAs levels until the end of culture. The PA increase in the diploid callus in the dark was related to free Put levels increase. The ploidy status of the plants could express different PA gradients together with the plant pith and in vitro callus cultures.

Differentiation of Melipona quadrifasciata L. (Hymenoptera, Apidae, Meliponini) subspecies using cytochrome b PCR-RFLP patterns

Melipona quadrifasciata quadrifasciata and M. quadrifasciata anthidioides are subspecies of M. quadrifasciata, a stingless bee species common in coastal Brazil. These subspecies are discriminated by the yellow stripe pattern of the abdominal tergites. We found Vsp I restriction patterns in the cytochrome b region closely associated to each subspecies in 155 M. quadrifasciata colonies of different geographical origin. This mitochondrial DNA molecular marker facilitates diagnosis of M. quadrifasciata subspecies matrilines and can be used to establish their natural distribution and identify hybrid colonies.

Hemodynamic, morphometric and autonomic patterns in hypertensive rats - renin-angiotensin system modulation

BACKGROUND: Spontaneously hypertensive rats develop left ventricular hypertrophy, increased blood pressure and blood pressure variability, which are important determinants of heart damage, like the activation of renin-angiotensin system. AIMS: To investigate the effects of the time-course of hypertension over 1) hemodynamic and autonomic patterns (blood pressure; blood pressure variability; heart rate); 2) left ventricular hypertrophy; and 3) local and systemic Renin-angiotensin system of the spontaneously hypertensive rats. METHODS: Male spontaneously hypertensive rats were randomized into two groups: young (n=13) and adult (n=12). Hemodynamic signals (blood pressure, heart rate), blood pressure variability (BPV) and spectral analysis of the autonomic components of blood pressure were analyzed. LEFT ventricular hypertrophy was measured by the ratio of LV mass to body weight (mg/g), by myocyte diameter (μm) and by relative fibrosis area (RFA, %). ACE and ACE2 activities were measured by fluorometry (UF/min), and plasma renin activity (PRA) was assessed by a radioimmunoassay (ng/mL/h). Cardiac gene expressions of Agt, Ace and Ace2 were quantified by RT-PCR (AU). RESULTS: The time-course of hypertension in spontaneously hypertensive rats increased BPV and reduced the alpha index in adult spontaneously hypertensive rats. Adult rats showed increases in left ventricular hypertrophy and in RFA. Compared to young spontaneously hypertensive rats, adult spontaneously hypertensive rats had lower cardiac ACE and ACE2 activities, and high levels of PRA. No change was observed in gene expression of Renin-angiotensin system components. CONCLUSIONS: The observed autonomic dysfunction and modulation of Renin-angiotensin system activity are contributing factors to end-organ damage in hypertension and could be interacting. Our findings suggest that the management of hypertensive disease must start before blood pressure reaches the highest stable levels and the consequent established end-organ damage is reached.