Generating Efficient Data Movement Code for Heterogeneous Architectures with Distributed-Memory

Programming for parallel architectures that do not have a shared address space is extremely difficult due to the need for explicit communication between memories of different compute devices. A heterogeneous system with CPUs and multiple GPUs, or a distributed-memory cluster are examples of such systems. Past works that try to automate data movement for distributed-memory architectures can lead to excessive redundant communication. In this paper, we propose an automatic data movement scheme that minimizes the volume of communication between compute devices in heterogeneous and distributed-memory systems. We show that by partitioning data dependences in a particular non-trivial way, one can generate data movement code that results in the minimum volume for a vast majority of cases. The techniques are applicable to any sequence of affine loop nests and works on top of any choice of loop transformations, parallelization, and computation placement. The data movement code generated minimizes the volume of communication for a particular configuration of these. We use a combination of powerful static analyses relying on the polyhedral compiler framework and lightweight runtime routines they generate, to build a source-to-source transformation tool that automatically generates communication code. We demonstrate that the tool is scalable and leads to substantial gains in efficiency. On a heterogeneous system, the communication volume is reduced by a factor of 11X to 83X over state-of-the-art, translating into a mean execution time speedup of 1.53X. On a distributed-memory cluster, our scheme reduces the communication volume by a factor of 1.4X to 63.5X over state-of-the-art, resulting in a mean speedup of 1.55X. In addition, our scheme yields a mean speedup of 2.19X over hand-optimized UPC codes.

A model experiment of the plate movement

Detailed analysis of some difficult aspects has been made from modeling the platemovement. A new method of using differential density of material (instead of differentialtemperature) has been developed in the experiments. The effect of convection of mantleon the plate movement has been studied using a centrifugal technique, and a patternshowing the recurrence of the plate movement has been successfully obtained. In this paper, a criterion De=Dm is presented for the similarity of the model to thecounterpart of the original mantle. According to the criterion, what happens in the modelin a span of ten minutes suggests a process of the "original model" going on in geologi-cal time of three million years.

Growth, movement, and attrition of northern bluefin tuna, Thunnus thynnus, in the Pacific Ocean, as determined by tagging

ENGLISH: The growth of northern bluefin tuna is described by a two-stanza model. For fish between 191 and 564 mm in length the Gompertz curve, with values of 581 mm and 4.32 for Loo and K (annual), respectively, is used. The fish between 564 and 1530 mm grow linearly, at the rate of 0.709 mm per day. Age-O fish tagged and released in the western Pacific Ocean have been recaptured in the western, central, and eastern Pacific. The minimum time between release in the western Pacific and recapture in the eastern Pacific is 215 days. Older fish, mostly Land 2-year olds, tagged and released in the eastern Pacific have been recaptured in the eastern and western Pacific. The minimum time between release in eastern Pacific and recapture in the western Pacific is 674 days. The coefficient of natural mortality is estimated from data on growth and ambient temperature to be 0.276 on an annual basis, with 90-percent confidence limits of 0.161 and 0.47L Spawning of northern bluefin takes place only in the western Pacific. Some of the juveniles migrate to the eastern Pacific, where they reside for several months to several years before returning to the western Pacific. The portion of fish which migrate to the eastern Pacific varies among years, and this appears to be an important cause of the annual variation in the catches in the eastern Pacific Ocean. SPANISH: El crecimiento del atún aleta azul del norte es descrito por un modelo de dos estadios. Para los peces de entre 191 y 564 mm de talla se usa la curva de Gompertz, con valores de 581 mm y 4.32 para Loo y K (anual), respectivamente. Los peces de entre 564 y 1530 mm crecen de forma lineal, a 0.709 mm por día. Peces de edad Omarcados y liberados en el Pacífico occidental han sido recapturados en el Pacífico occidental, central, y oriental. La demora mínima entre la liberación en el Pacífico occidental y la recaptura en el Pacífico oriental es de 215 días. Peces mayores, principalmente de 1 ó 2 años de edad, marcados y liberados en el Pacífico oriental han sido re capturados en el Pacífico occidental y oriental. La demora mínima entre la liberación en el Pacífico oriental y la recaptura en el Pacífico occidental es de 674 días. Se estima el coeficiente de mortalidad natural a partir de los datos de crecimiento y temperatura ambiental en un 0.276 anual, con límites de confianza al 90% de 0.161 y 0.471. El aleta azul del norte desova únicamente en el Pacífico occidental. Algunos de los juveniles migran al Pacífico oriental, donde permanecen entre varios meses y varios años antes de regresar al Pacífico occidental. La porción de los peces que migran al Pacífico oriental varía entre años, y ésto parece ser una causa importante de la variación anual en las capturas en el Océano Pacífico oriental. (PDF contains 94 pages.)

The oculomotor system: (1) Vertical-horizontal interaction and signal recognition, (2) Time delays and power spectra

In the first section of this thesis, two-dimensional properties of the human eye movement control system were studied. The vertical - horizontal interaction was investigated by using a two-dimensional target motion consisting of a sinusoid in one of the directions vertical or horizontal, and low-pass filtered Gaussian random motion of variable bandwidth (and hence information content) in the orthogonal direction. It was found that the random motion reduced the efficiency of the sinusoidal tracking. However, the sinusoidal tracking was only slightly dependent on the bandwidth of the random motion. Thus the system should be thought of as consisting of two independent channels with a small amount of mutual cross-talk.

These target motions were then rotated to discover whether or not the system is capable of recognizing the two-component nature of the target motion. That is, the sinusoid was presented along an oblique line (neither vertical nor horizontal) with the random motion orthogonal to it. The system did not simply track the vertical and horizontal components of motion, but rotated its frame of reference so that its two tracking channels coincided with the directions of the two target motion components. This recognition occurred even when the two orthogonal motions were both random, but with different bandwidths.

In the second section, time delays, prediction and power spectra were examined. Time delays were calculated in response to various periodic signals, various bandwidths of narrow-band Gaussian random motions and sinusoids. It was demonstrated that prediction occurred only when the target motion was periodic, and only if the harmonic content was such that the signal was sufficiently narrow-band. It appears as if general periodic motions are split into predictive and non-predictive components.

For unpredictable motions, the relationship between the time delay and the average speed of the retinal image was linear. Based on this I proposed a model explaining the time delays for both random and periodic motions. My experiments did not prove that the system is sampled data, or that it is continuous. However, the model can be interpreted as representative of a sample data system whose sample interval is a function of the target motion.

It was shown that increasing the bandwidth of the low-pass filtered Gaussian random motion resulted in an increase of the eye movement bandwidth. Some properties of the eyeball-muscle dynamics and the extraocular muscle "active state tension" were derived.

The content of the form of Julia Alvarezs In the time of the butterflies

In the Time of the Butterflies é um romance da escritora dominicana-americana Julia Alvarez sobre a vida e a morte das Borboletas, Las Mariposas, codinome das irmãs Mirabal, membros de um movimento clandestino contra o regime ditatorial de Rafael Leonidas Trujillo na República Dominicana, que se tornaram símbolos da luta contra o Trujillato depois de serem assassinadas a mando do ditador. Essa dissertação tem como objetivo expor como forma literária e contexto social estão diretamente relacionados nesse romance. Ela defende a ideia de que o borramento de três gêneros literários distintos metaficção historiográfica, autobiografia e bildungsroman reflete o questionamento das fronteiras entre o privado e o público, o pessoal e o político, o eu e o outro, o individual e o coletivo, a literatura e a história, fato e ficção e história e subjetividade. Ela também tenta mostrar como a problematização dessas dicotomias implica na contestação de noções pré-concebidas de identidade, história e nação

Age-specific movement patterns of sablefish (Anoplopoma fimbria) in Alaska

Over 34,000 age 0–2 juvenile sablefish (Anoplopoma fimbria) were tagged and released in southeast Alaska waters during 1985–2005. The data set resulting from this tagging study was unusual because of its time span (20 years) and because age could be reliably inferred from release length (i.e., tagged and released fish were of known age); thus, age-specific movement patterns could be examined. The depth- and area-related recovery patterns supported the concepts that sablefish move to deeper water with age and migrate counterclockwise in the Gulf of Alaska. Availability to the fishery increased rapidly for fish of younger ages, peaked at age 5 to 6, and then gradually declined as sablefish moved deeper with age. Decreased availability with age may occur because of lower fishing effort in deep water and could have substantial implications for sablef ish stock assessments because “domeshaped” availability influences the reliability of abundance estimates. The area-related recovery pattern was not affected by year-class strength; i.e., there was no significant densitydependent relationship.

Vertical movement patterns of skipjack tuna (Katsuwonus pelamis) in the eastern equatorial Pacific Ocean, as revealed with archival tags

Thirty-three skipjack tuna (Katsuwonus pelamis) (53−73 cm fork length) were caught and released with implanted archival tags in the eastern equatorial Pacific Ocean during April 2004. Six skipjack tuna were recap-tured, and 9.3 to 10.1 days of depth and temperature data were down-loaded from five recovered tags. The vertical habitat-use distributions indicated that skipjack tuna not associated with floating objects spent 98.6% of their time above the thermocline (depth=44 m) during the night, but spent 37.7% of their time below the thermocline during the day. When not associated with floating objects, skipjack tuna displayed repetitive bounce-diving behavior to depths between 50 and 300 m during the day. The deepest dive recorded was 596 m, where the ambient temperature was 7.7°C. One dive was particularly remarkable because the fish contin-uously swam for 2 hours below the thermocline to a maximum depth of 330 m. During that dive, the ambient temperature reached a low of 10.5°C, and the peritoneal cavity temperature reached a low of 15.9°C. The vertical movements and habitat use of skipjack tuna, revealed in this study, provide a much greater understanding of their ecological niche and catchability by purse-seine fisheries.

Transboundary Movement of Atlantic Istiophorid Billfishes Among International and U.S. Domestic Management Areas Inferred from Mark-Recapture Studies

Billfish movements relative to the International Commission for the Conservation of Atlantic Tunas management areas, as well as U.S. domestic data collection areas within the western North Atlantic basin, were investigated with mark-recapture data from 769 blue marlin, Makaira nigricans, 961 white marlin, Tetrapturus albidus, and 1,801 sailfish, Istiophorus platypterus. Linear displacement between release and recapture locations ranged from zero (all species) to 15,744 km (mean 575, median 119, SE 44) for blue marlin, 6,523 km (mean 719, median 216, SE 33) for white marlin, and 3,845 km (mean 294, median 98, SE 13) for sailfish. In total, 2,824 (80.0%) billfish were recaptured in the same management area of release. Days at liberty ranged from zero (all species) to 4,591 (mean 619, median 409, SE 24) for blue marlin, 5,488 (mean 692, median 448, SE 22) for white marlin, and 6,568 (mean 404, median 320, SE 11) for sailfish. The proportions (per species) of visits were highest in the Caribbean area for blue marlin and white marlin, and the Florida East Coast area for sailfish. Blue marlin and sailfish were nearly identical when comparing the percent of individuals vs. the number of areas visited. Overall, white marlin visited more areas than either blue marlin or sailfish. Seasonality was evident for all species, with overall results generally reflecting the efforts of the catch and release recreational fishing sector, particularly in the western North Atlantic. This information may be practical in reducing the uncertainties in billfish stock assessments and may offer valuable insight into management consideration of time-area closure regulations to reduce bycatch mortality of Atlantic billfishes.

Near optimal combination of sensory and motor uncertainty in time during a naturalistic perception-action task.

Most behavioral tasks have time constraints for successful completion, such as catching a ball in flight. Many of these tasks require trading off the time allocated to perception and action, especially when only one of the two is possible at any time. In general, the longer we perceive, the smaller the uncertainty in perceptual estimates. However, a longer perception phase leaves less time for action, which results in less precise movements. Here we examine subjects catching a virtual ball. Critically, as soon as subjects began to move, the ball became invisible. We study how subjects trade-off sensory and movement uncertainty by deciding when to initiate their actions. We formulate this task in a probabilistic framework and show that subjects' decisions when to start moving are statistically near optimal given their individual sensory and motor uncertainties. Moreover, we accurately predict individual subject's task performance. Thus we show that subjects in a natural task are quantitatively aware of how sensory and motor variability depend on time and act so as to minimize overall task variability.

Mortality and movement of Yellowtail Flounder (Limanda ferruginea) tagged off New England

From 2003 to 2006, 44,882 Yellowtail Flounder (Limanda ferruginea) were captured and released with conventional disc tags in the western North Atlantic as part of a cooperative Yellowtail Flounder tagging study. From these releases, 3767 of the tags were recovered. The primary objectives of this tagging program were to evaluate the mortality and large-scale movement of Yellowtail Flounder among 3 stock areas in New England. To explore mortality, survival and recovery rate were estimated from traditional Brownie tag-recovery models fitted to the data with Program MARK. Models were examined with time and sex-dependent parameters over several temporal scales. The models with a monthly scale for both survival and recovery rate had the best overall fit and returned parameter estimates that were biologically reasonable. Estimates of survival from the tag-recovery models confirm the general magnitude of total mortality derived from age-based stock assessments but indicate that survival was greater for females than for males. In addition to calculating mortality estimates, we examined the pattern of release and recapture locations and revealed frequent movements within stock areas and less frequent movement among stock areas. The collaboration of fishermen and scientists for this study successfully resulted in independent confirmation of previously documented patterns of movement and mortality rates from conventional age-based analyses.

Fish movement patterns in Virgin Islands National Park, Virgin Islands Coral Reef National Monument and adjacent waters

NOAA’s National Centers for Coastal Ocean Science (NCCOS)-Center for Coastal Monitoring and Assessment’s (CCMA) Biogeography Branch, National Park Service (NPS), US Geological Survey, and the University of Hawaii used acoustic telemetry to quantify spatial patterns and habitat affinities of reef fishes around the island of St. John, US Virgin Islands. The objective of the study was to define the movements of reef fishes among habitats within and between the Virgin Islands Coral Reef National Monument (VICRNM), the Virgin Islands National Park (VIIS), and Territorial waters surrounding St. John. In order to better understand species’ habitat utilization patterns among management regimes, we deployed an array of hydroacoustic receivers and acoustically tagged reef fishes. Thirty six receivers were deployed in shallow near-shore bays and across the shelf to depths of approximately 30 m. One hundred eighty four individual fishes were tagged representing 19 species from 10 different families with VEMCO V9-2L-R64K transmitters. The array provides fish movement information at fine (e.g., day-night and 100s meters within a bay) to broad spatial and temporal scales (multiple years and 1000s meters across the shelf). The long term multi-year tracking project provides direct evidence of connectivity across habitat types in the seascape and among management units. An important finding for management was that a number of individuals moved among management units (VICRNM, VINP, Territorial waters) and several snapper moved from near-shore protected areas to offshore shelf-edge spawning aggregations. However, most individuals spent the majority of their time with VIIS and VICRNM, with only a few wide-ranging species moving outside the management units. Five species of snappers (Lutjanidae) accounted for 31% of all individuals tagged, followed by three species of grunts (Haemulidae) accounting for an additional 23% of the total. No other family had more than a single species represented in the study. Bluestripe grunt (Haemulon sciurus) comprised 22% of all individuals tagged, followed by lane snappers (Lutjanus synagris) at 21%, bar jack (Carangoides ruber) at 11%, and saucereye porgy (Calamus calamus) at 10%. The largest individual tagged was a 70 cm TL nurse shark (Ginglymostoma cirratum), followed by a 65 cm mutton snapper (Lutjanus analis), a 47 cm bar jack, and a 41 cm dog snapper (Lutjanus jocu). The smallest individuals tagged were a 19 cm blue tang (Acanthurus coeruleus) and a 19.2 cm doctorfish (Acanthurus chirurgus). Of the 40 bluestriped grunt acoustically tagged, 73% were detected on the receiver array. The average days at large (DAL) was 249 (just over 8 months), with one individual detected for 930 days (over two and a half years). Lane snapper were the next most abundant species tagged (N = 38) with 89% detected on the array. The average days at large (DAL) was 221 with one individual detected for 351 days. Seventy-one percent of the bar jacks (N = 21) were detected on the array with the average DALs at 47 days. All of the mutton snapper (N = 12) were detected on the array with an average DAL of 273 and the longest at 784. The average maximum distance travelled (MDT) was ca. 2 km with large variations among species. Grunts, snappers, jacks, and porgies showed the greatest movements. Among all individuals across species, there was a positive and significant correlation between size of individuals and MDT and between DAL and MDT.

Preliminary results from the Caribbean Acoustic Tracking Network (CATn): a data sharing partnership for acoustic tracking and movement ecology of marine animals in the Caribbean Sea

We have recently exchanged and integrated into a single database tag detections for conch, teleost and elasmobranch fish from four separately maintained arrays in the U.S. Virgin Islands including the NMFS queen conch array (St. John nearshore), NOAA’s Biogeography Branch array (St. John nearshore & midshelf reef); UVI shelf edge arrays (Marine Conservation District, Grammanik & other shelf edge); NOAA NMFS Apex Predator array COASTSPAN (St. John nearshore). The integrated database has over 7.5 million hits. Data is shared only with consent of partners and full acknowledgements. Thus, the summary of integrated data here uses data from NOAA and UVI arrays under a cooperative agreement. The benefits of combining and sharing data have included increasing the total area of detection resulting in an understanding of broader scale connectivity than would have been possible with a single array. Partnering has also been cost-effectiveness through sharing of field work, staff time and equipment and exchanges of knowledge and experience across the network. Use of multiple arrays has also helped in optimizing the design of arrays when additional receivers are deployed. The combined arrays have made the USVI network one of the most extensive acoustic arrays in the world with a total of 150+ receivers available, although not necessarily all deployed at all times. Currently, two UVI graduate student projects are using acoustic array data.

Individual growth rates and movement of juvenile white shrimp (Litopenaeus setiferus) in a tidal marsh nursery

We measured growth and movements of individually marked free-ranging juvenile white shrimp (Litopenaeus setiferus) in tidal creek subsystems of the Duplin River, Sapelo Island, Georgia. Over a period of two years, 15,974 juvenile shrimp (40−80 mm TL) were marked internally with uniquely coded microwire tags and released in the shallow upper reaches of four salt marsh tidal creeks. Subsequent samples were taken every 3−6 days from channel segments arranged at 200-m intervals along transects extending from the upper to lower reach of each tidal creek. These collections included 201,384 juvenile shrimp, of which 184 were marked recaptures. Recaptured shrimp were at large an average of 3−4 weeks (range: 2−99 days) and were recovered a mean distance of <0.4 km from where they were initially marked. Mean residence times in the creek subsystems ranged from 15.2 to 25.5 days and were estimated from exponential decay functions describing the proportions of marked individuals recaptured with increasing days at large. Residence time was not significantly correlated with creek length (Pearson=−0.316, P=0.684 ), but there was suggestive evidence of positive associations with either intertidal (Pearson r=0.867, P=0.133) or subtidal (Pearson r=0.946, P=0.054) drainage area. Daily mean specific growth rates averaged 0.009 to 0.013 among creeks; mean absolute growth rates ranged from 0.56−0.84 mm/d, and were lower than those previously reported for juvenile penaeids in estuaries of the southeastern United States. Mean individual growth rates were not significantly different between years (t-test, P>0.30) but varied significantly during the season, tending to be greater in July than November. Growth rates were size-dependent, and temporal changes in size distributions rather than temporal variation in physical environmental factors may have accounted for seasonal differences in growth. Growth rates differed between creeks in 1999 (t-test, P<0.015), but not in 1998 (t-test, P>0.5). We suggest that spatial variation in landscape structure associated with access to intertidal resources may have accounted for this apparent interannual difference in growth response.

On the domain decomposition and transmission line modelling finite element method for time-domain induction motor analysis

This paper demonstrates how a finite element model which exploits domain decomposition is applied to the analysis of three-phase induction motors. It is shown that a significant gain in cpu time results when compared with standard finite element analysis. Aspects of the application of the method which are particular to induction motors are considered: the means of improving the convergence of the nonlinear finite element equations; the choice of symmetrical sub-domains; the modelling of relative movement; and the inclusion of periodic boundary conditions. © 1999 IEEE.

Impedance control is selectively tuned to multiple directions of movement

Humans are able to learn tool-handling tasks, such as carving, demonstrating their competency to make movements in unstable environments with varied directions. When faced with a single direction of instability, humans learn to selectively co-contract their arm muscles tuning the mechanical stiffness of the limb end point to stabilize movements. This study examines, for the first time, subjects simultaneously adapting to two distinct directions of instability, a situation that may typically occur when using tools. Subjects learned to perform reaching movements in two directions, each of which had lateral instability requiring control of impedance. The subjects were able to adapt to these unstable interactions and switch between movements in the two directions; they did so by learning to selectively control the end-point stiffness counteracting the environmental instability without superfluous stiffness in other directions. This finding demonstrates that the central nervous system can simultaneously tune the mechanical impedance of the limbs to multiple movements by learning movement-specific solutions. Furthermore, it suggests that the impedance controller learns as a function of the state of the arm rather than a general strategy. © 2011 the American Physiological Society.