Quantum computers are increasingly being integrated into classical computing environments, particularly within supercomputing and data centers, where they serve as accelerators for solving a wide range of specific problems. Previous research has indicated that quantum computers have the potential to resolve optimization problems with theoretically lower computational complexity than classical algorithms. However, in practical scenarios, even minor instances of these problems often require substantial computational resources, enhanced quality quantum hardware, and the results may not always be optimal. In this paper, we highlight the importance of hybrid systems, combining quantum and classical computations to address growing problem sizes and computational demands. We focus on optimization problems using hardware and software-enhanced quantum computers operational within a novel quantum–classical hybrid setup, incorporating GPUs and photonic quantum computers. The study specifically addresses combinatorial optimization problems, such as the Max-Cut and the Job Shop Scheduling Problem, in addition to selected machine learning classification use cases. Notably, the quantum algorithm in Max-Cut optimization outperformed complete solution searches, especially for larger problem instances. For the Job-Shop Scheduling Problem, we made a significant advancement by successfully solving substantially larger instances compared to our previous previous work. Furthermore, selected hybrid neural networks incorporating quantum layers showed improved stability, though without a clear quality advantage over classical models. The paper also highlights the rapid progress and technological achievements in both hardware and software used in near-term photonic quantum computers, suggesting a promising future for quantum–classical hybrid systems in useful applications.

One of the most critical application areas for many operational High-Performance Computing systems worldwide is a numerical weather prediction. Such complex predictions and corresponding computing models have to solve a large number of Partial Differential Equations using stencil computations on structured grids within tight production schedules. A stencil kernel within simulations is often the most demanding computing part and naturally may impact the energy consumption of the whole HPC system. In this paper, we introduce new functional extensions developed for DCworms and GSSIM simulators to predict the energy efficiency of extreme-scale stencil applications on heterogeneous computing resources, in particular consisting of a large number of many-core and GPUs. New energy-efficiency metrics for modeling of heterogeneous computing resources have been successfully added to the simulators thanks to their pluggable and extensible architectures. Our recent improvements related to application-specific performance models in simulators help users to take into account many relevant application parameters, in particular, those related to detailed energy consumption on heterogeneous HPC resources. We show in this paper how to extract stencil application-specific parameters based on real experiments. Moreover, we demonstrate how those parameters can be applied for modeling and simulation experiments to evaluate the overall performance and energy consumption of stencil computations on manycore CPUs and GPUs. Finally, we discuss new DCworms simulator capabilities and demonstrate added-values of performance analysis tools.

Grid simulation tools provide frameworks for simulating application scheduling in various Grid infrastructures. However, while experimenting with many existing tools, we have encountered two main shortcomings: (i) there are no tools for generating workloads, resources and events ; (ii) it is difficult and time consuming to model different Grid levels, i.e. resource brokers, and local level scheduling systems. In this paper we present the Grid Scheduling Simulator (GSSIM), a framework that addresses these shortcomings and provides an easy-to-use Grid scheduling framework for enabling simulations of a wide range of scheduling algorithms in multi-level, heterogeneous Grid infrastructures. In order to foster more collaboration in the community at large, GSSIM is complemented with a portal (http://www.gssim.org) that provides a repository of Grid scheduling algorithms, synthetic workloads and benchmarks for use with GSSIM.

The problem of optimal scheduling n independent malleable tasks in a parallel processor system is studied. It is assumed that an execution of any task can be preempted and the number of processors allocated to the same task can change during its execution. We present a rectangle packing algorithm, which converts an optimal solution for the relaxed problem, in which the number of processors allocated to a task is not required to be integer, into an optimal solution for the original problem in O(n) time.

MOTIVATION: Reconstruction of the original DNA sequence in the sequencing by the hybridization approach (SBH) requires computational support due to a large number of possible combinations. One can notice a lack of algorithms admitting false-negative data and giving in addition all possible solutions. RESULTS: In this paper, a new method of sequencing has been proposed. An algorithm based on its idea (for the general case, when some data are missing, like in the real experiment) has been implemented and tested. Authentic DNA sequences have been used for testing. A parallel version of the algorithm has also been implemented and tested. The quality of the reconstruction is satisfactory for the library of oligonucleotides of length between 8 and 12, and 100, 200 and 300 bp long sequences. A way to a further decrease in the computation time is also suggested.

The problem considered in this paper is the deterministic scheduling of tasks on a set of identical processors. However, the model presented differs from the classical one by the requirement that certain tasks need more than one processor at a time for their processing. This assumption is especially justified in some microprocessor applications and its impact on the complexity of minimizing schedule length is studied. First we concentrate on the problem of nonpreemptive scheduling. In this case, polynomial-time algorithms exist only for unit processing times. We present two such algorithms of complexity O(n) for scheduling tasks requiring an arbitrary number of processors between 1 and k at a time where k is a fixed integer. The case for which k is not fixed is shown to be NP-complete. Next, the problem of preemptive scheduling of tasks of arbitrary length is studied. First an algorithm for scheduling tasks requiring one or k processors is presented. Its complexity depends linearly on the number of tasks. Then, the possibility of a linear programming formulation for the general case is analyzed.

This paper proposes a deterministic approach to the preemptive scheduling of independent tasks, which takes into account primary memory allocation in multiprocessor systems with virtual memory and a common primary memory. Each central processing unit (CPU) is assumed to have dedicated paging devices and thus paging- device competition does not exist in the system. The system workload is based on an analytic approximation to the lifetime curve of a task. Exact and approximate algorithms are presented which minimize or tend to minimize the length of schedules on an arbitrary number of identical processors. In the general case, the exact algorithm is based on nonlinear programming; however, the approximate algorithm requires the solution of several nonlinear equations with one unknown. For certain cases, analytical results have also been obtained.

Allocation is discussed of constrained resources among activities of a network project, when the resource requirements of activities concern a unit of a discrete resource (machine, processor) from a finite set of m parallel units and an amount of a continuously divisible resource (power, fuel flow, approximate manpower) which is arbitrary within a certain interval. For every activity the function relating the performing speed to the allotted amount of continuous resource is known as is the state which has to be reached in order to complete the activity. Two optimality criteria; project duration and the mean finishing time of an activity are considered. For the first criterion the way in which finding the optimal solution is reduced to a constrained nonlinear programming problem is described. The number of variables in this problem depends on the number of m-element combinations of activities which may be performed simultaneously in accordance with the precedence constraints. Consequently, this approach is of more theoretical than practical importance. For some special cases, however, it allows analytical results to be obtained. Next, an approximate method is described which consists of two stages. In the first stage the problem of scheduling activities with known performing times on parallel machines is solved, and in the second, the continuous resource is allocated among the activities (or parts of activities) which are performed simultaneously in the obtained schedule.

PurposeSubroutine A R S M E solves a resource constrained, network scheduling problem for the case in which activities may be arbitrarily interrupted and restarted later with no increase in activity duration.The number of resource types is not a limiting factor in our procedure.The amount of any one resource available at any moment is constant.We shall use the "activity-on-arc" network representation, under the commonly imposed assumption that the network contains no directed cycles and has only one "beginning" and only one "terminal" node (event).I t is further assumed that the network nodes (events) are ordered in such a way t h a t node i precedes node j, if i < j.Such an ordering is always possible and it induces an ordering among the arcs (activities).Optimal approaches to resource constrained, network scheduling problems where activities can require more than one resource type are presented in [1,4].Both methods assume integer durations of activities, and the method presented in [1] divides activity durations into unit intervals.Both methods can handle networks with up to about 30 activities and 3 resource types.Subroutine A R S M E is constructed in such a way that its storage requirements are minimal, a fact which permits the solution of problems for very large networks with many resource types.Moreover, an optimal solution can be obtained in a shorter time when relatively smaller amounts of the resources are available than when resources are less limited.Let the number of activities be equal to M and the number of resource types be equal to RT.For activityj (j = 1, 2 , . . ., M) and resource k (k = 1, 2 , . . ., RT)