Graphics Processing Units Processing Research Articles

Real-time parallel 3D multiple particle tracking with single molecule centrifugal force microscopy.

Real-time tracking of multiple particles is key for quantitative analysis of dynamic biophysical processes and materials science via time-lapse microscopy image data, especially for single molecule biophysical techniques, such as magnetic tweezers and centrifugal force microscopy. However, real-time multiple particle tracking with high resolution is limited by the current imaging processes or tracking algorithms. Here, we demonstrate 1 nm resolution in three dimensions in real-time with a graphics-processing unit (GPU) based on a compute unified device architecture (CUDA) parallel computing framework instead of only a central processing unit (CPU). We also explore the trade-offs between processing speed and size of the utilized regions of interest and a maximum speedup of 137 is achieved with the GPU compared with the CPU. Moreover, we utilize this method with our recently self-built centrifugal force microscope (CFM) in experiments that track multiple DNA-tethered particles. Our approach paves the way for high-throughput single molecule techniques with high resolution and efficiency. LAY DESCRIPTION: Particles are widely used as probes in life sciences through their motions. In single molecule techniques such as optical tweezers and magnetic tweezers, microbeads are used to study intermolecular or intramolecular interactions via beads tracking. Also tracking multiple beads' motions could study cell-cell or cell-ECM interactions in traction force microscopy. Therefore, particle tracking is of key important during these researches. However, parallel 3D multiple particle tracking in real-time with high resolution is a challenge either due to the algorithm or the program. Here, we combine the performance of CPU and CUDA-based GPU to make a hybrid implementation for particle tracking. In this way, a speedup of 137 is obtained compared the program before only with CPU without loss of accuracy. Moreover, we improve and build a new centrifugal force microscope for multiple single molecule force spectroscopy research in parallel. Then we employed our program into centrifugal force microscope for DNA stretching study. Our results not only demonstrate the application of this program in single molecule techniques, also indicate the capability of multiple single molecule study with centrifugal force microscopy.

A Power-Performance Perspective to Multiobjective Electroencephalogram Feature Selection on Heterogeneous Parallel Platforms.

This article provides an insight on the power-performance issues related with the CPU-GPU (Central Processing Unit-Graphics Processing Unit) parallel implementations of problems that frequently appear in the context of applications on bioinformatics and biomedical engineering. More specifically, we analyze the power-performance behavior of an evolutionary parallel multiobjective electroencephalogram feature selection procedure that evolves subpopulations of solutions with time-demanding fitness evaluation. The procedure has been implemented in OpenMP to dynamically distribute either subpopulations or individuals among devices, and uses OpenCL to evaluate the fitness of the individuals. The development of parallel codes usually implies to maximize the code efficiency, thus optimizing the achieved speedups. To follow the same trend, this article extends and provides a more complete analysis of our previous works about the power-performance characteristics in heterogeneous CPU-GPU platforms considering different operation frequencies and evolutionary parameters, such as distribution of individuals, etc. This way, different experimental configurations of the proposed procedure have been evaluated and compared with respect to a master-worker approach, not only in runtime but also considering energy consumption. The experimental results show that lower operating frequencies does not necessarily mean lower energy consumptions since energy is the product of power and time. Thus, we have observed that parallel processing not only reduces the runtime, but also the energy consumed by the application despite a higher instantaneous power. Particularly, the workload distribution among both CPU and GPU cores provides the best runtime and very low energy consumption compared with the values achieved by the same alternatives executed by only CPU threads.

A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations

The parallel computation capabilities of modern GPU (Graphics Processing Unit) processors have attracted increasing attention from researchers and engineers who have been conducting high computational throughput studies. However, current single GPU based engineering solutions are often struggle to fulfill their real-time requirements. Thus, the multi-GPU-based approach has become a popular and cost-effective choice for tackling the demands. In those cases, the computational load balancing over multiple GPU “nodes” is often the key and bottleneck that affect the quality and performance of the runtime system. The existing load balancing approaches are mainly based on the assumption that all GPU nodes in the same computer framework are of equal computational performance, which are often not the case due to cluster design and other legacy issues. This paper presents a novel dynamic load balancing (DLB) model for rapid data division and allocation on heterogeneous GPU nodes based on an innovative fuzzy neural network (FNN). In this research, a 5-state parameter feedback mechanism defining the overall cluster and node performances is proposed. The corresponding FNN-based DLB model will be capable of monitoring and predicting individual node performance under different workload scenarios. A real-time adaptive scheduler has been devised to reorganize the data inputs to each node when necessary to maintain their runtime computational performances. The devised model has been implemented on two dimensional (2D) discrete wavelet transform (DWT) tasks for evaluation. Experiment results show that this DLB model has enabled a high computational throughput while ensuring real-time and precision requirements from complex computational tasks.

Simulating heterogeneous behaviours in complex systems on GPUs

Agent Based Modelling (ABM) is an approach for modelling dynamic systems and studying complex and emergent behaviour. ABMs have been widely applied in diverse disciplines including biology, economics, and social sciences. The scalability of ABM simulations is typically limited due to the computationally expensive nature of simulating a large number of individuals. As such, large scale ABM simulations are excellent candidates to apply parallel computing approaches such as Graphics Processing Units (GPUs). In this paper, we present an extension to the FLAME GPU11Flexible Large-scale Agent-based Modelling simulation platform on GPU [1] framework which addresses the divergence problem, i.e. the challenge of executing the behaviour of non-homogeneous individuals on vectorised GPU processors. We do this by describing a modelling methodology which exposes inherent parallelism within the model which is exploited by novel additions to the software permitting higher levels of concurrent simulation execution. Moreover, we demonstrate how this extension can be applied to realistic cellular level tissue model by benchmarking the model to demonstrate a measured speedup of over 4x.

PAGANI Toolkit: Parallel graph‐theoretical analysis package for brain network big data

The recent collection of unprecedented quantities of neuroimaging data with high spatial resolution has led to brain network big data. However, a toolkit for fast and scalable computational solutions is still lacking. Here, we developed the PArallel Graph-theoretical ANalysIs (PAGANI) Toolkit based on a hybrid central processing unit-graphics processing unit (CPU-GPU) framework with a graphical user interface to facilitate the mapping and characterization of high-resolution brain networks. Specifically, the toolkit provides flexible parameters for users to customize computations of graph metrics in brain network analyses. As an empirical example, the PAGANI Toolkit was applied to individual voxel-based brain networks with ∼200,000 nodes that were derived from a resting-state fMRI dataset of 624 healthy young adults from the Human Connectome Project. Using a personal computer, this toolbox completed all computations in ∼27 h for one subject, which is markedly less than the 118 h required with a single-thread implementation. The voxel-based functional brain networks exhibited prominent small-world characteristics and densely connected hubs, which were mainly located in the medial and lateral fronto-parietal cortices. Moreover, the female group had significantly higher modularity and nodal betweenness centrality mainly in the medial/lateral fronto-parietal and occipital cortices than the male group. Significant correlations between the intelligence quotient and nodal metrics were also observed in several frontal regions. Collectively, the PAGANI Toolkit shows high computational performance and good scalability for analyzing connectome big data and provides a friendly interface without the complicated configuration of computing environments, thereby facilitating high-resolution connectomics research in health and disease.

Connectivity derived thalamic segmentation in deep brain stimulation for tremor.

The ventral intermediate nucleus (VIM) of the thalamus is an established surgical target for stereotactic ablation and deep brain stimulation (DBS) in the treatment of tremor in Parkinson's disease (PD) and essential tremor (ET). It is centrally placed on a cerebello-thalamo-cortical network connecting the primary motor cortex, to the dentate nucleus of the contralateral cerebellum through the dentato-rubro-thalamic tract (DRT). The VIM is not readily visible on conventional MR imaging, so identifying the surgical target traditionally involved indirect targeting that relies on atlas-defined coordinates. Unfortunately, this approach does not fully account for individual variability and requires surgery to be performed with the patient awake to allow for intraoperative targeting confirmation. The aim of this study is to identify the VIM and the DRT using probabilistic tractography in patients that will undergo thalamic DBS for tremor. Four male patients with tremor dominant PD and five patients (three female) with ET underwent high angular resolution diffusion imaging (HARDI) (128 diffusion directions, 1.5 mm isotropic voxels and b value = 1500) preoperatively. Patients received VIM-DBS using an MR image guided and MR image verified approach with indirect targeting. Postoperatively, using parallel Graphical Processing Unit (GPU) processing, thalamic areas with the highest diffusion connectivity to the primary motor area (M1), supplementary motor area (SMA), primary sensory area (S1) and contralateral dentate nucleus were identified. Additionally, volume of tissue activation (VTA) corresponding to active DBS contacts were modelled. Response to treatment was defined as 40% reduction in the total Fahn-Tolosa-Martin Tremor Rating Score (FTMTRS) with DBS-ON, one year from surgery. Three out of nine patients had a suboptimal, long-term response to treatment. The segmented thalamic areas corresponded well to anatomically known counterparts in the ventrolateral (VL) and ventroposterior (VP) thalamus. The dentate-thalamic area, lay within the M1-thalamic area in a ventral and lateral location. Streamlines corresponding to the DRT connected M1 to the contralateral dentate nucleus via the dentate-thalamic area, clearly crossing the midline in the mesencephalon. Good response was seen when the active contact VTA was in the thalamic area with highest connectivity to the contralateral dentate nucleus. Non-responders had active contact VTAs outside the dentate-thalamic area. We conclude that probabilistic tractography techniques can be used to segment the VL and VP thalamus based on cortical and cerebellar connectivity. The thalamic area, best representing the VIM, is connected to the contralateral dentate cerebellar nucleus. Connectivity based segmentation of the VIM can be achieved in individual patients in a clinically feasible timescale, using HARDI and high performance computing with parallel GPU processing. This same technique can map out the DRT tract with clear mesencephalic crossing.

Parallelization and Performance of the NIM Weather Model on CPU, GPU, and MIC Processors

Abstract The design and performance of the Non-Hydrostatic Icosahedral Model (NIM) global weather prediction model is described. NIM is a dynamical core designed to run on central processing unit (CPU), graphics processing unit (GPU), and Many Integrated Core (MIC) processors. It demonstrates efficient parallel performance and scalability to tens of thousands of compute nodes and has been an effective way to make comparisons between traditional CPU and emerging fine-grain processors. The design of the NIM also serves as a useful guide in the fine-grain parallelization of the finite volume cubed (FV3) model recently chosen by the National Weather Service (NWS) to become its next operational global weather prediction model. This paper describes the code structure and parallelization of NIM using standards-compliant open multiprocessing (OpenMP) and open accelerator (OpenACC) directives. NIM uses the directives to support a single, performance-portable code that runs on CPU, GPU, and MIC systems. Performance results are compared for five generations of computer chips including the recently released Intel Knights Landing and NVIDIA Pascal chips. Single and multinode performance and scalability is also shown, along with a cost–benefit comparison based on vendor list prices.

Priority-driven spatial resource sharing scheduling for embedded graphics processing units

Many visual tasks in modern personal devices such smartphones resort heavily to graphics processing units (GPUs) for their fluent user experiences. Because most GPUs for embedded systems are non-preemptive by nature, it is important to schedule GPU resources efficiently across multiple GPU tasks. We present a novel spatial resource sharing (SRS) technique for GPU tasks, called a budget-reservation spatial resource sharing (BR-SRS) scheduling, which limits the number of GPU processing cores for a job based on the priority of the job. Such a priority-driven resource assignment can prevent a high-priority foreground GPU task from being delayed by background GPU tasks. The BR-SRS scheduler is invoked only twice at the arrival and completion of jobs, and thus, the scheduling overhead is minimized as well. We evaluated the performance of our scheduling scheme in an Android-based smartphone, and found that the proposed technique significantly improved the performance of high-priority tasks in comparison to the previous temporal budget-based multi-task scheduling.

CENTRAL PROCESSING UNIT-GRAPHICS PROCESSING UNIT COMPUTING SCHEME FOR MULTI-OBJECT TRACKING IN SURVEILLANCE

This research work presents a novel central processing unit-graphics processing unit (CPU-GPU) computing scheme for multiple object trackingduring a surveillance operation. This facilitates nonlinear computational jobs to avail completion of computation in minimal processing time for tracking function. The work is divided into two essential objectives. First is to dynamically divide the processing operations into parallel units, and second is to reduce the communication between CPU-GPU processing units.

A Load-Aware Pluggable Cloud Framework for Real-Time Video Processing

A large number of video applications require real-time response. The high-speed video processing then requires a distributed and parallelized framework utilizing all possible computing resources, i.e., both Central Processing Unit (CPU) and Graphics Processing Unit (GPU) at their best. The CPU–GPU collaboration may cause resource imbalance where GPU-based jobs consume less computing resources while occupying more memory compared with CPU-based jobs. In this paper, we propose a load-aware pluggable cloud framework for real-time video processing where CPU–GPU switching based on workload status can be performed at runtime. Furthermore, we design aspect-oriented monitors to collect framework metrics and propose a distance coverage algorithm to detect performance degradation in order to make sure that the framework runs optimally to achieve good performance when a load-aware task switching is made. We have comprehensively evaluated the framework and the evaluation results show that the proposed framework has good performance, reusability, pluggability, and scalability.

Graph-Based Simultaneous Localization and Mapping: Computational Complexity Reduction on a Multicore Heterogeneous Architecture

This article deals with the computational complexity issue of graphbased simultaneous localization and mapping (SLAM). SLAM allows a robot that is navigating in an unknown environment to build a map of this environment while simultaneously determining the robot pose on this map. Graph-based SLAM is a smoothing method that uses a graph to represent and solve the SLAM problem. We first propose a graph construction that takes advantage of the incremental and sparse characteristics of graph-based SLAM. This incremental construction is exploited to perform several algorithmic optimizations. Second, we present a study of using a heterogeneous architecture to implement the graph-based SLAM algorithm. Indeed, the emergence of recent heterogeneous embedded architectures should lead to a great advance in the design of embedded systems-based robotics applications. As a result of this study, an algorithm-architecture mapping is proposed for a central processing unit-graphics processing unit (CPU-GPU)-based architecture. The study also investigates how this kind of architecture can speed up graph-based SLAM by offloading some critical compute-intensive tasks of the algorithm on the GPU. Some common data sets are used to compare our implementations to the state of the art.

Step Ring-Based Three-Dimensional Path Planning Via Graphics Processing Unit Simulation for Subtractive Three-Dimensional Printing

In this paper, both software model visualization with path simulation and associated machining product are produced based on the step ring-based three-axis path planning to demo model-driven graphics processing unit (GPU) feature in tool path planning and 3D image model classification by GPU simulation. Subtractive 3D printing (i.e., 3D machining) is represented as integration between 3D printing modeling and computer numerical control (CNC) machining via GPU simulated software. Path planning is applied through visualization of surface material removal in high-resolution and 3D path simulation via ring selective path planning based on accessibility of path through pattern selection. First, the step ring selects critical features to reconstruct computer-aided design (CAD) design model as stereolithography (STL) voxel, and then, local optimization is attained within interested ring area for time and energy saving of GPU volume generation as compared to global automatic path planning with longer latency. The reconstructed CAD model comes from an original sample (GATech buzz) with 2D image information. CAD model for optimization and validation is adopted to sustain manufacturing reproduction based on system simulation feedback. To avoid collision with the produced path from retraction path, we pick adaptive ring path generation and prediction in each planning iteration, which may also minimize material removal. Moreover, we did partition analysis and G-code optimization for large-scale model and high density volume data. Image classification and grid analysis based on adaptive 3D tree depth are proposed for multilevel set partition of the model to define no cutting zones. After that, accessibility map is computed based on accessibility space for rotational angular space of path orientation to compare step ring-based pass planning verses global path planning of all geometries. Feature analysis via central processing unit (CPU) or GPU processor for GPU map computation contributes to high-performance computing and cloud computing potential through parallel computing application of subtractive 3D printing in the future.

Performance analysis of finite-difference time-domain schemes for acoustic simulation implemented on multi-core and many-core processor architectures

There have been many discussions about hardware-accelerations for the large-scale finite-difference time-domain (FDTD) method to deal with various wave dynamics in scientific and engineering computations. Recently, there is growing interest in many-core based parallel computing with graphics processing unit (GPU) or many integrated Core (MIC) accelerators. Therefore, it is worth applying those to large-scale and long-time acoustic simulation for FDTD methods. In this paper, performance analyses are performed for three types of acoustic FDTD schemes, which are FDTD(2,2), FDTD(2,4) and wave equation FDTD (WE-FDTD)(2,2), implemented on GPU, Intel MIC, and multi-core central processing unit (CPU). Here, FDTD(T, S) denotes Tth-order accuracy for the time derivative and Sth-order accuracy for the spatial derivative, respectively. The parallelized FDTD schemes are implemented with Compute Unified Device Architecture (CUDA) and OpenACC, respectively, on GPU, and OpenMP for MIC and multi-core CPU. As a method for performance analysis, we employ the modified roofline model which adds the concept of cache hit ratio to operational intensity. The analysis provides performance comparison results of attainable performance and measured performance for each FDTD scheme and each processor. We performed software optimizations for MIC based on the performance analysis and achieved 78%-91% of attainable performance on MIC.

Parallel algorithms constructing the cell graph

SummaryMotion 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.

Video Anomaly Detection in Real Time on a Power-Aware Heterogeneous Platform

Field-programmable gate arrays (FPGAs) and graphics processing units (GPUs) are often used when real-time performance in video processing is required. An accelerated processor is chosen based on task-specific priorities (power consumption, processing time, and detection accuracy), and this decision is normally made once at design time. All three of the characteristics are important, particularly in battery-powered systems. Here, we propose a method for moving selection of processing platform from a single design-time choice to a continuous run-time one. We implement Histogram of Oriented Gradients (HOG) for cars and people and Mixture-of-Gaussians motion detectors running across FPGA, GPU, and central processing unit in a heterogeneous system. We use this to detect illegally parked vehicles in urban scenes. Power, time, and accuracy information for each detector is characterized. An anomaly measure is assigned to each detected object based on its trajectory and location compared with learned contextual movement patterns. This drives processor and implementation selection so that scenes with high behavioral anomalies are processed with faster but more power-hungry implementations, but routine or static time periods are processed with power-optimized and less accurate slower versions. Real-time performance is evaluated on video data sets including i-LIDS. Compared with power-optimized static selection, automatic dynamic implementation mapping is 10% more accurate, but draws 12-W of extra power in our testbed desktop system.

Big Data Approaches for the Analysis of Large-Scale fMRI Data Using Apache Spark and GPU Processing: A Demonstration on Resting-State fMRI Data from the Human Connectome Project.

Technologies for scalable analysis of very large datasets have emerged in the domain of internet computing, but are still rarely used in neuroimaging despite the existence of data and research questions in need of efficient computation tools especially in fMRI. In this work, we present software tools for the application of Apache Spark and Graphics Processing Units (GPUs) to neuroimaging datasets, in particular providing distributed file input for 4D NIfTI fMRI datasets in Scala for use in an Apache Spark environment. Examples for using this Big Data platform in graph analysis of fMRI datasets are shown to illustrate how processing pipelines employing it can be developed. With more tools for the convenient integration of neuroimaging file formats and typical processing steps, big data technologies could find wider endorsement in the community, leading to a range of potentially useful applications especially in view of the current collaborative creation of a wealth of large data repositories including thousands of individual fMRI datasets.

Efficient methods for implementation of multi-level nonrigid mass-preserving image registration on GPUs and multi-threaded CPUs

Background and objectiveFaster and more accurate methods for registration of images are important for research involved in conducting population-based studies that utilize medical imaging, as well as improvements for use in clinical applications. We present a novel computation- and memory-efficient multi-level method on graphics processing units (GPU) for performing registration of two computed tomography (CT) volumetric lung images. MethodsWe developed a computation- and memory-efficient Diffeomorphic Multi-level B-Spline Transform Composite (DMTC) method to implement nonrigid mass-preserving registration of two CT lung images on GPU. The framework consists of a hierarchy of B-Spline control grids of increasing resolution. A similarity criterion known as the sum of squared tissue volume difference (SSTVD) was adopted to preserve lung tissue mass. The use of SSTVD consists of the calculation of the tissue volume, the Jacobian, and their derivatives, which makes its implementation on GPU challenging due to memory constraints. The use of the DMTC method enabled reduced computation and memory storage of variables with minimal communication between GPU and Central Processing Unit (CPU) due to ability to pre-compute values. The method was assessed on six healthy human subjects. ResultsResultant GPU-generated displacement fields were compared against the previously validated CPU counterpart fields, showing good agreement with an average normalized root mean square error (nRMS) of 0.044±0.015. Runtime and performance speedup are compared between single-threaded CPU, multi-threaded CPU, and GPU algorithms. Best performance speedup occurs at the highest resolution in the GPU implementation for the SSTVD cost and cost gradient computations, with a speedup of 112 times that of the single-threaded CPU version and 11 times over the twelve-threaded version when considering average time per iteration using a Nvidia Tesla K20X GPU. ConclusionsThe proposed GPU-based DMTC method outperforms its multi-threaded CPU version in terms of runtime. Total registration time reduced runtime to 2.9min on the GPU version, compared to 12.8min on twelve-threaded CPU version and 112.5min on a single-threaded CPU. Furthermore, the GPU implementation discussed in this work can be adapted for use of other cost functions that require calculation of the first derivatives.

Reverse time migration of 3D vertical seismic profile data

Reverse time migration (RTM) has shown increasing advantages in handling seismic images of complex subsurface media, but it has not been used widely in 3D seismic data due to the large storage and computation requirements. Our prime objective was to develop an RTM strategy that was applicable to 3D vertical seismic profiling (VSP) data. The strategy consists of two aspects: storage saving and calculation acceleration. First, we determined the use of the random boundary condition (RBC) to save the storage in wavefield simulation. An absorbing boundary such as the perfect matching layer boundary is often used in RTM, but it has a high memory demand for storing the source wavefield. RBC is a nonabsorbing boundary and only stores the source wavefield at the two maximum time steps, then repropagates the source wavefield backwards at every time step, and hence, it significantly reduces the memory requirement. Second, we examined the use of the graphic processing unit (GPU) parallelization technique to accelerate the computation. RBC needs to simulate the source wavefield twice and doubles the computation. Thus, it is very necessary to realize the RTM algorithm by GPU, especially for a 3D VSP data set. GPU and central processing unit (CPU) collaborated parallel implementation can greatly reduce the computation time, where the CPU performs serial code, and the GPU performs parallel code. Because RBC does not need the same huge amount of storage as an absorbing boundary, RTM becomes practically applicable for 3D VSP imaging.

Real-time GPU-accelerated processing and volumetric display for wide-field laser-scanning optical-resolution photoacoustic microscopy.

Fast signal processing and real-time displays are essential for practical imaging modality in various fields of applications. However, the imaging speed in optical-resolution photoacoustic microscopy (OR-PAM), in particular, depends on factors such as the pulse repetition rate of the laser, scanning method, field of view (FOV), and signal processing time. In the past, efforts to increase acquisition speed either focused on developing new scanning methods or using lasers with higher pulse repetition rates. However, high-speed signal processing is also important for real-time volumetric display in OR-PAM. In this study, we carried out parallel signal processing using a graphics processing unit (GPU) to enable fast signal processing and wide-field real-time displays in laser-scanning OR-PAM. The average total GPU processing time for a B-mode PAM image was approximately 1.35 ms at a display speed of 480 fps when the data samples were acquired with 736 (axial) × 500 (lateral) points/B-mode-frame at a pulse repetition rate of 300 kHz. In addition, we successfully displayed maximum amplitude projection images of a mouse's ear as volumetric images with an FOV of 3 mm × 3 mm (500 × 500 pixels) at 1.02 s, corresponding to 0.98 fps.

GPU accelerated parallel FFT processing for Fourier transform hyperspectral imaging

Sequentially processing massive 1D fast Fourier transformations (FFT) on raw interferograms using a CPU has limited the speed of conventional Fourier transform imaging spectrometers (FTIS). This paper reports the implementation of highly paralleled FFT computation using a low-end graphics processing unit (GPU) device for acceleration of this process. Comparison experiment results have demonstrated ∼10–30 times acceleration improvement using GPU-based parallel processing over conventional CPU-based serial processing upon the input data with same sizes: GPU processing time of only 630 and 173 ms of datacubes with 512×512×1024 and 64×64×16 k pixels, respectively, has presented its potential for online and even real-time FTIS raw data processing. The addition of a cheap GPU device into any FTIS system involves no optical modifications, so it is a highly cost-effective technique for temporal resolution enhancement of FTIS-based hyperspectral imaging applications.