Abstract: In a system for improving performance of tensor-based computations and for minimizing the associated memory usage, computations associated with different non-zero tensor values are performed while exploiting an overlap between the respective index tuples of those non-zero values. While performing computations associated with a selected mode, when an index corresponding to a particular mode in a current index tuple is the same as the corresponding index from another, previously processed index tuple, the value already stored in a buffer corresponding to that particular mode is reused either wholly or in part, minimizing the processor usage and improving performance. Certain matrix operations may be iterated more than once so as to avoid the need to store a large partial result obtained from those operations. The performance overhead of the repeated operations is not significant, but the reduction in memory usage is.

Abstract: A system for extracting latent information from data includes obtaining or generating components of the data, where the data components include scores indicating how the component relates to the data. Memory is allocated for the components and the components are stored in the allocated memory. The components are then transformed into documents using a suitable transformation function, and the documents are analyzed using natural language processing, to extract latent information contained in the data.

Abstract: A system for allocation of one or more data structures used in a program across a number of processing units takes into account a memory access pattern of the data structure, and the amount of total memory available for duplication across the several processing units. Using these parameters duplication factors are determined for the one or more data structures such that the cost of remote communication is minimized when the data structures are duplicated according to the respective duplication factors while allowing parallel execution of the program.

Abstract: In a network system, an application receiving packets can consume one or more packets in two or more stages, where the second and the later stages can selectively consume some but not all of the packets consumed by the preceding stage. Packets are transferred between two consecutive stages, called producer and consumer, via a fixed-size storage. Both the producer and the consumer can access the storage without locking it and, to facilitate selective consumption of the packets by the consumer, the consumer can transition between awake and sleep modes, where the packets are consumed in the awake mode only. The producer may also switch between awake and sleep modes. Lockless access is made possible by controlling the operation of the storage by the producer and the consumer both according to the mode of the consumer, which is communicated via a shared memory location.

Abstract: In a system for storing in memory a tensor that includes at least three modes, elements of the tensor are stored in a mode-based order for improving locality of references when the elements are accessed during an operation on the tensor. To facilitate efficient data reuse in a tensor transform that includes several iterations, on a tensor that includes at least three modes, a system performs a first iteration that includes a first operation on the tensor to obtain a first intermediate result, and the first intermediate result includes a first intermediate-tensor. The first intermediate result is stored in memory, and a second iteration is performed in which a second operation on the first intermediate result accessed from the memory is performed, so as to avoid a third operation, that would be required if the first intermediate result were not accessed from the memory.

Abstract: In a system for efficiently detecting large/elephant flows in a network, the rate at which the received packets are sampled is adjusted according to a top flow detection likelihood computed for a cache of flows identified in the arriving network traffic. After observing packets sampled from the network, Dirichlet-Categorical inference is employed to calculate a posterior distribution that captures uncertainty about the sizes of each flow, yielding a top flow detection likelihood. The posterior distribution is used to find the most likely subset of elephant flows. The technique rapidly converges to the optimal sampling rate at a speed O(1/n), where n is the number of packet samples received, and the only hyperparameter required is the targeted detection likelihood.

Abstract: We present the architecture of a high-performance constraint solver R-Solve that extends the gains made in SAT performance over the past fifteen years on static decision problems to problems that require on-the-fly adaptation, solution space exploration and optimization. R-Solve facilitates collaborative parallel solving and provides an efficient system for unrestricted incremental solving via Smart Repair. R-Solve can address problems in dynamic planning and constrained optimization involving complex logical and arithmetic constraints.

Abstract: In a system for enabling configuration of an ensemble of several solvers, such that the ensemble can efficiently solve a constraint problem, for each one of several candidate configurations, an array of scores is computed. The array corresponds to a statistical parameter related to a problem solution, and the computation is based on, at least in part, a set of features associated with the problem. One candidate configuration is assigned to a solver, and based on the array of scores associated with that candidate configuration the same or a different candidate configuration is assigned to a another solver. A system for dynamically reconfiguring an ensemble of solvers obtains runtime data from several solvers, and a new configuration is determined by applying a machine learning and/or heuristic analysis procedure to the runtime data. The configuration of a solver may be updated according to the new configuration while that solver is running.

Abstract: A system for performing tensor decomposition in a selective expansive and/or recursive manner, a tensor is decomposed into a specified number of components, and one or more tensor components are selected for further decomposition. For each selected component, the significant elements thereof are identified, and using the indices of the significant elements a sub-tensor is formed. In a subsequent iteration, each sub-tensor is decomposed into a respective specified number of components. Additional sub-tensors corresponding to the components generated in the subsequent iteration are formed, and these additional sub-tensors may be decomposed further in yet another iteration, until no additional components are selected. The mode of a sub-tensor can be decreased or increased prior to decomposition thereof. Components likely to reveal information about the data stored in the tensor can be selected for decomposition.

Abstract: A system for compiling programs for execution thereof using a hierarchical processing system having two or more levels of memory hierarchy can perform memory-level-specific optimizations, without exceeding a specified maximum compilation time. To this end, the compiler system employs a polyhedral model and limits the dimensions of a polyhedral program representation that is processed by the compiler at each level using a focalization operator that temporarily reduces one or more dimensions of the polyhedral representation. Semantic correctness is provided via a defocalization operator that can restore all polyhedral dimensions that had been temporarily removed.

Abstract: In a sequence of major computational steps or in an iterative computation, a stencil amplifier can increase the number of data elements accessed from one or more data structures in a single major step or iteration, thereby decreasing the total number of computations and/or communication operations in the overall sequence or the iterative computation. Stencil amplification, which can be optimized according to a specified parameter such as compile time, run time, code size, etc., can improve the performance of a computing system executing the sequence or the iterative computation in terms of run time, memory load, energy consumption, etc. The stencil amplifier typically determines boundaries, to avoid erroneously accessing data elements not present in the one or more data structures.

Abstract: Methods, apparatus and computer software product for optimization of data transfer between two memories includes determining access to master data stored in one memory and/or to local data stored in another memory such that either or both of the size of total data transferred and the number of data transfers required to transfer the total data can be minimized. The master and/or local accesses are based on, at least in part, respective structures of the master and local data.

Abstract: In a system for automatic generation of event-driven, tuple-space based programs from a sequential specification, a hierarchical mapping solution can target different runtimes relying on event-driven tasks (EDTs). The solution uses loop types to encode short, transitive relations among EDTs that can be evaluated efficiently at runtime. Specifically, permutable loops translate immediately into conservative point-to-point synchronizations of distance one. A runtime-agnostic which can be used to target the transformed code to different runtimes.

Abstract: A compilation system can define, at compile time, the data blocks to be managed by an Even Driven Task (EDT) based runtime/platform, and can also guide the runtime/platform on when to create and/or destroy the data blocks, so as to improve the performance of the runtime/platform. The compilation system can also guide, at compile time, how different tasks may access the data blocks they need in a manner that can improve performance of the tasks.

Abstract: A compilation system using an energy model based on a set of generic and practical hardware and software parameters is presented. The model can represent the major trends in energy consumption spanning potential hardware configurations using only parameters available at compilation time. Experimental verification indicates that the model is nimble yet sufficiently precise, allowing efficient selection of one or more parameters of a target computing system so as to minimize power/energy consumption of a program while achieving other performance related goals. A voltage and/or frequency optimization and selection is presented which can determine an efficient dynamic hardware configuration schedule at compilation time. In various embodiments, the configuration schedule is chosen based on its predicted effect on energy consumption. A concurrency throttling technique based on the energy model can exploit the power-gating features exposed by the target computing system to increase the energy efficiency of programs.

Abstract: A system for allocation of one or more data structures used in a program across a number of processing units takes into account a memory access pattern of the data structure, and the amount of total memory available for duplication across the several processing units. Using these parameters duplication factors are determined for the one or more data structures such that the cost of remote communication is minimized when the data structures are duplicated according to the respective duplication factors while allowing parallel execution of the program.

Abstract: A system for determining the physical path of an object can map several candidate paths to a suitable path space that can be explored using a convex optimization technique. The optimization technique may take advantage of the typical sparsity of the path space and can identify a likely physical path using a function of sensor observation as constraints. A track of an object can also be determined using a track model and a convex optimization technique.

Abstract: A system for detecting and/or tracking a moving object uses signals received from the object at two or more receivers, each of which is dual polarized. At each receiver, the component of the received signal associated with the polarization thereof is separated from the non-polarization-based component, and the modulation of the polarization-based component is used for locating and/or tracking the moving object.

Abstract: We present the architecture of a high-performance constraint solver R-Solve that extends the gains made in SAT performance over the past fifteen years on static decision problems to problems that require on-the-fly adaptation, solution space exploration and optimization. R-Solve facilitates collaborative parallel solving and provides an efficient system for unrestricted incremental solving via Smart Repair. R-Solve can address problems in dynamic planning and constrained optimization involving complex logical and arithmetic constraints.

Abstract: A multiresolution parser (MRP) can selectively extract one or more information units from a dataset based on the available processing capacity and/or the arrival rate of the dataset. Should any of these parameters change, the MRP can adaptively change the information units to be extracted such that the benefit or value of the extracted information is maximized while minimizing the cost of extraction. This tradeoff is facilitated, at least in part, by an analysis of the spectral energy of the datasets expected to be processed by the MRP. The MRP can also determine its state after a processing iteration and use that state information in subsequent iterations to minimize the required computations in such subsequent iterations, so as to improve processing efficiency.