GPU clusters underpin modern deep learning, yet studies across industry and academia consistently report widespread GPU underutilization. Prior work and our own analysis indicate that inefficiency often stems from recurring patterns in code, job scripts, and runtime behaviour that users rarely detect. We argue that addressing this issue is a MSR challenge: it requires mining inefficiency patterns, combining static and dynamic signals for actionable feedback, validating job-submission artefacts, and developing privacy-aware datasets linking code, configuration, and runtime metrics.

MOTIVATION: Investigating the 3D structure of chromatin provides new insights into transcriptional regulation. With the evolution of 3C next-generation sequencing methods like ChiA-PET and Hi-C, the surge in data volume has highlighted the need for more efficient chromatin spatial modelling algorithms. This study introduces the cudaMMC method, based on the Simulated Annealing Monte Carlo approach and enhanced by GPU-accelerated computing, to efficiently generate ensembles of chromatin 3D structures. RESULTS: The cudaMMC calculations demonstrate significantly faster performance with better stability compared to our previous method on the same workstation. cudaMMC also substantially reduces the computation time required for generating ensembles of large chromatin models, making it an invaluable tool for studying chromatin spatial conformation. AVAILABILITY AND IMPLEMENTATION: Open-source software and manual and sample data are freely available on https://github.com/SFGLab/cudaMMC.

We demonstrate a new idea of a parallel GPU JSON parser, which is able to optimize the parsing and initial transformation process through metaprogramming. It outperforms other well-known solutions like simdjson, Pandas, as well as cuDF– which also works on GPU. The resulting data is ready to be further processed in common data frame formats and may be incorporated by RAPIDS, Apache Arrow or Pandas. Our parser can therefore be a part of an industrial Extract-Transform-Load workflow.

In order to speed up the process of rendering scenes containing many light sources, spatial data structures are used, which allow the number of lights processed for each pixel to be reduced during lighting computation.Examples of algorithms using such data structures are clustered shading and hybrid lighting.Alongside the rendering time, it is important to consider memory consumption resulting from processing a large number of lights.This paper presents a novel modification of the hybrid lighting algorithm using an octree that allows for a significant reduction in the amount of memory required to store the data structure.The proposed modification uses an octree to store the information about the rendered space.Detailed analysis of the proposed algorithm, and numerical results obtained for various 3D scenes, as well as different input data, all prove that the proposed method significantly reduces the memory required to store lists of lights used by the algorithm.

Summary R‐Trie dictionary is a multi‐way retrieval tree for arbitrary length binary keys. It is designed taking into account utilization of thousands of parallel vector threads, a fundamental technique of programming GPU processors. Dictionary bulk creation algorithm automatically scales to use as many threads as elements in the input batch, without any synchronization or page locking, exceeding 20·10 6 key insertions per second on a consumer class device. Bulk search procedure achieves 600·10 6 lookups per second in case of a Longest Prefix Match algorithm and 3·10 9 lookups per second in case of exact key matching. In this paper, we describe details of implementation, present results of run‐time experiments, and enumerate several open problems for future.

Summary Motion planning is an important and well‐studied field of robotics. A typical approach to finding a route is to construct a cell graph representing a scene and then to find a path in such a graph. In this paper, we present and analyze several parallel algorithms for constructing the cell graph on a single instruction, multiple data‐like graphics processing unit (GPU) processor. GPU utilization is necessary because of insufficient processing power of CPUs reported by other authors. A GPU processor with its parallel processing capabilities promises some improvement if only proper implementations of the algorithms can be found. We show that a naive brute force algorithm, enhanced by a simple heuristics, in an average case, can be faster than comprehensive solutions based on parallel implementation of an asymptotically optimal sequential algorithm. Copyright © 2016 John Wiley & Sons, Ltd.

Graphics Processing Units (GPU) have significantly more applications than just rendering images. They are also used in general-purpose computing to solve problems that can benefit from massive parallel processing. However, there are tasks that either hardly suit GPU or fit GPU only partially. The latter class is the focus of this paper. We elaborate on hybrid CPU/GPU computation and build optimization methods that seek the equilibrium between these two computation platforms. The method is based on heuristic search for bi-objective Pareto optimal execution plans in presence of multiple concurrent queries. The underlying model mimics the commodity market where devices are producers and queries are consumers. The value of resources of computing devices is controlled by supply-and-demand laws. Our model of the optimization criteria allows finding solutions of problems not yet addressed in heterogeneous query processing. Furthermore, it also offers lower time complexity and higher accuracy than other methods.