multi-GPU Systems Research Articles

Optimization of machine learning model training procedure on multi-gpu systems to enhance cyber security in telecommunication networks

This paper analyzes the optimization features of machine learning (ML) model training procedures using multi-GPU systems to enhance cyber security in telecommunication networks. A key aspect of the study is the use of data parallelism, which allows the distribution of the training load across multiple GPUs, significantly reducing training time and improving model accuracy-critical factors for rapid threat detection in cyberspace. A novel approach for optimizing data batch size using Mutual Information (MI) is proposed, which harmonizes the utilization of computational resources with the information content of the training data. MI helps to determine the optimal data batch size that minimizes training errors and improves model accuracy without a significant increase in training time. Experimental results demonstrate the substantial advantages of multi-GPU configurations compared to single-GPU setups, providing faster training and improved model accuracy. It was particularly emphasized that MI-guided batch size tuning significantly outperforms traditional manual tuning methods, ensuring higher validation accuracy and reducing training time. The study showed that the MI-based approach is an effective tool for addressing the problem of optimizing ML model training processes in real-world scenarios where cyber security is critical. The proposed methods allow ML models to train faster and more accurately identify potential threats, making them particularly relevant for telecommunication networks where a rapid response to new threats in real time is required. The implementation of modern computational technologies such as multi-GPU systems and MI-optimized training enhances the efficiency and accuracy of machine learning models. This, in turn, improves cyber security measures and ensures a more reliable defence of telecommunication networks against malicious attacks. It is noted that the proposed approaches can be adapted not only for cyber security but also for other areas where high model accuracy and fast training are important. Future research prospects include the development of new machine learning methods, particularly deep neural networks, the exploration of alternative computational architectures such as quantum computing or distributed systems, and their integration into real-time systems. Special attention should be paid to the ethical aspects of implementing automated cyber security systems, particularly in preventing bias in algorithms and ensuring fairness in their application.

Laminography as a tool for imaging large-size samples with high resolution.

Despite the increased brilliance of the new generation synchrotron sources, there is still a challenge with high-resolution scanning of very thick and absorbing samples, such as a whole mouse brain stained with heavy elements, and, extending further, brains of primates. Samples are typically cut into smaller parts, to ensure a sufficient X-ray transmission, and scanned separately. Compared with the standard tomography setup where the sample would be cut into many pillars, the laminographic geometry operates with slab-shaped sections significantly reducing the number of sample parts to be prepared, the cutting damage and data stitching problems. In this work, a laminography pipeline for imaging large samples (>1 cm) at micrometre resolution is presented. The implementation includes a low-cost instrument setup installed at the 2-BM micro-CT beamline of the Advanced Photon Source. Additionally, sample mounting, scanning techniques, data stitching procedures, a fast reconstruction algorithm with low computational complexity, and accelerated reconstruction on multi-GPU systems for processing large-scale datasets are presented. The applicability of the whole laminography pipeline was demonstrated by imaging four sequential slabs throughout an entire mouse brain sample stained with osmium, in total generating approximately 12 TB of raw data for reconstruction.

Adaptively Placed Multi-Grid Scene Representation Networks for Large-Scale Data Visualization.

Scene representation networks (SRNs) have been recently proposed for compression and visualization of scientific data. However, state-of-the-art SRNs do not adapt the allocation of available network parameters to the complex features found in scientific data, leading to a loss in reconstruction quality. We address this shortcoming with an adaptively placed multi-grid SRN (APMGSRN) and propose a domain decomposition training and inference technique for accelerated parallel training on multi-GPU systems. We also release an open-source neural volume rendering application that allows plug-and-play rendering with any PyTorch-based SRN. Our proposed APMGSRN architecture uses multiple spatially adaptive feature grids that learn where to be placed within the domain to dynamically allocate more neural network resources where error is high in the volume, improving state-of-the-art reconstruction accuracy of SRNs for scientific data without requiring expensive octree refining, pruning, and traversal like previous adaptive models. In our domain decomposition approach for representing large-scale data, we train an set of APMGSRNs in parallel on separate bricks of the volume to reduce training time while avoiding overhead necessary for an out-of-core solution for volumes too large to fit in GPU memory. After training, the lightweight SRNs are used for realtime neural volume rendering in our open-source renderer, where arbitrary view angles and transfer functions can be explored. A copy of this paper, all code, all models used in our experiments, and all supplemental materials and videos are available at https://github.com/skywolf829/APMGSRN.

PARALiA: A Performance Aware Runtime for Auto-tuning Linear Algebra on Heterogeneous Systems

Dense linear algebra operations appear very frequently in high-performance computing (HPC) applications, rendering their performance crucial to achieve optimal scalability. As many modern HPC clusters contain multi-GPU nodes, BLAS operations are frequently offloaded on GPUs, necessitating the use of optimized libraries to ensure good performance. Unfortunately, multi-GPU systems are accompanied by two significant optimization challenges: data transfer bottlenecks as well as problem splitting and scheduling in multiple workers (GPUs) with distinct memories. We demonstrate that the current multi-GPU BLAS methods for tackling these challenges target very specific problem and data characteristics, resulting in serious performance degradation for any slightly deviating workload. Additionally, an even more critical decision is omitted because it cannot be addressed using current scheduler-based approaches: the determination of which devices should be used for a certain routine invocation. To address these issues we propose a model-based approach: using performance estimation to provide problem-specific autotuning during runtime. We integrate this autotuning into an end-to-end BLAS framework named PARALiA. This framework couples autotuning with an optimized task scheduler, leading to near-optimal data distribution and performance-aware resource utilization. We evaluate PARALiA in an HPC testbed with 8 NVIDIA-V100 GPUs, improving the average performance of GEMM by 1.7× and energy efficiency by 2.5× over the state-of-the-art in a large and diverse dataset and demonstrating the adaptability of our performance-aware approach to future heterogeneous systems.

Exploring the Formation of Resistive Pseudodisks with the GPU Code Astaroth

Pseudodisks are dense structures formed perpendicular to the direction of the magnetic field during the gravitational collapse of a molecular cloud core. Numerical simulations of the formation of pseudodisks are usually computationally expensive with conventional CPU codes. To demonstrate the proof of concept of a fast computing method for this numerically costly problem, we explore the GPU-powered MHD code Astaroth, a sixth-order finite difference method with low adjustable finite resistivity implemented with sink particles. The formation of pseudodisks is physically and numerically robust and can be achieved with a simple and clean setup for this newly adopted numerical approach for science verification. The method’s potential is illustrated by evidencing the dependence on the initial magnetic field strength of specific physical features accompanying the formation of pseudodisks, e.g., the occurrence of infall shocks and the variable behavior of the mass and magnetic flux accreted on the central object. As a performance test, we measure both weak and strong scaling of our implementation to find the most efficient way to use the code on a multi-GPU system. Once suitable physics and problem-specific implementations are realized, the GPU-accelerated code is an efficient option for 3D magnetized collapse problems.

Distributed out-of-memory NMF on CPU/GPU architectures

AbstractWe propose an efficient distributed out-of-memory implementation of the non-negative matrix factorization (NMF) algorithm for heterogeneous high-performance-computing systems. The proposed implementation is based on prior work on NMFk, which can perform automatic model selection and extract latent variables and patterns from data. In this work, we extend NMFk by adding support for dense and sparse matrix operation on multi-node, multi-GPU systems. The resulting algorithm is optimized for out-of-memory problems where the memory required to factorize a given matrix is greater than the available GPU memory. Memory complexity is reduced by batching/tiling strategies, and sparse and dense matrix operations are significantly accelerated with GPU cores (or tensor cores when available). Input/output latency associated with batch copies between host and device is hidden using CUDA streams to overlap data transfers and compute asynchronously, and latency associated with collective communications (both intra-node and inter-node) is reduced using optimized NVIDIA Collective Communication Library (NCCL) based communicators. Benchmark results show significant improvement, from 32X to 76x speedup, with the new implementation using GPUs over the CPU-based NMFk. Good weak scaling was demonstrated on up to 4096 multi-GPU cluster nodes with approximately 25,000 GPUs when decomposing a dense 340 Terabyte-size matrix and an 11 Exabyte-size sparse matrix of density $$10^{-6}$$ 10 - 6 .

Effect-based Multi-viewer Caching for Cloud-native Rendering

With cloud computing becoming ubiquitous, it appears as virtually everything can be offered as-a-service. However, real-time rendering in the cloud forms a notable exception, where the cloud adoption stops at running individual game instances in compute centers. In this paper, we explore whether a cloud-native rendering architecture is viable and scales to multi-client rendering scenarios. To this end, we propose world-space and on-surface caches to share rendering computations among viewers placed in the same virtual world. We discuss how caches can be utilized on an effect-basis and demonstrate that a large amount of computations can be saved as the number of viewers in a scene increases. Caches can easily be set up for various effects, including ambient occlusion, direct illumination, and diffuse global illumination. Our results underline that the image quality using cached rendering is on par with screen-space rendering and due to its simplicity and inherent coherence, cached rendering may even have advantages in single viewer setups. Analyzing the runtime and communication costs, we show that cached rendering is already viable in multi-GPU systems. Building on top of our research, cloud-native rendering may be just around the corner.

Accelerating 4D image reconstruction for magnetic resonance-guided radiotherapy.

Physiological motion impacts the dose delivered to tumours and vital organs in external beam radiotherapy and particularly in particle therapy. The excellent soft-tissue demarcation of 4D magnetic resonance imaging (4D-MRI) could inform on intra-fractional motion, but long image reconstruction times hinder its use in online treatment adaptation. Here we employ techniques from high-performance computing to reduce 4D-MRI reconstruction times below two minutes to facilitate their use in MR-guided radiotherapy. Four patients with pancreatic adenocarcinoma were scanned with a radial stack-of-stars gradient echo sequence on a 1.5T MR-Linac. Fast parallelised open-source implementations of the extra-dimensional golden-angle radial sparse parallel algorithm were developed for central processing unit (CPU) and graphics processing unit (GPU) architectures. We assessed the impact of architecture, oversampling and respiratory binning strategy on 4D-MRI reconstruction time and compared images using the structural similarity (SSIM) index against a MATLAB reference implementation. Scaling and bottlenecks for the different architectures were studied using multi-GPU systems. All reconstructed 4D-MRI were identical to the reference implementation (SSIM 0.99). Images reconstructed with overlapping respiratory bins were sharper at the cost of longer reconstruction times. The CPU +GPU implementation was over 17 times faster than the reference implementation, reconstructing images in 60 1 s and hyper-scaled using multiple GPUs. Respiratory-resolved 4D-MRI reconstruction times can be reduced using high-performance computing methods for online workflows in MR-guided radiotherapy with potential applications in particle therapy.

Warp-Aware Adaptive Energy Efficiency Calibration for Multi-GPU Systems

Massive GPU acceleration processors have been used in high-performance computing systems. The Dennard-scaling has led to power and thermal constraints limiting the performance of such systems. The demand for both increased performance and energy-efficiency is highly desired. This paper presents a multi-layer low-power optimisation method for warps and tasks parallelisms. We present a dynamic frequency regulation scheme for performance parameters in terms of load balance and load imbalance. The method monitors the energy parameters in runtime and adjusts adaptively the voltage level to ensure the performance efficiency with energy reduction. The experimental results show that the multi-layer low-power optimisation with dynamic frequency regulation can achieve 40% energy consumption reduction with only 1.6% performance degradation, thus reducing 59% maximum energy consumption. It can further save about 30% energy consumption in comparison with the single-layer energy optimisation.

Enable high-resolution, real-time ensemble simulation and data assimilation of flood inundation using distributed GPU parallelization

Numerical modeling of the intensity and evolution of flood events are affected by multiple sources of uncertainty such as precipitation and land surface conditions. To quantify and curb these uncertainties, an ensemble-based simulation and data assimilation model for pluvial flood inundation is constructed. The shallow water equation is decoupled in the x and y directions, and the inertial form of the Saint-Venant equation is chosen to realize fast computation. The probability distribution of the input and output factors is described using Monte Carlo samples. Subsequently, a particle filter is incorporated to enable the assimilation of hydrological observations and improve prediction accuracy. To achieve high-resolution, real-time ensemble simulation, heterogeneous computing technologies based on CUDA (compute unified device architecture) and a distributed storage multi-GPU (graphics processing unit) system are used. Multiple optimization skills are employed to ensure the parallel efficiency and scalability of the simulation program. Taking an urban area of Fuzhou, China as an example, a model with a 3-m spatial resolution and 4.0 million units is constructed, and 8 Tesla P100 GPUs are used for the parallel calculation of 96 model instances. Under these settings, the ensemble simulation of a 1-hour hydraulic process takes 2.0 min, which achieves a 2680 × estimated speedup compared with a single-thread run on CPU. The calculation results indicate that the particle filter method effectively constrains simulation uncertainty while providing the confidence intervals of key hydrological elements such as streamflow, submerged area, and submerged water depth. The presented approaches show promising capabilities in handling the uncertainties in flood modeling as well as enhancing prediction efficiency.

Towards the Simulation of a Realistic Large-Scale Spiking Network on a Desktop Multi-GPU System.

The reproduction of the brain ’sactivity and its functionality is the main goal of modern neuroscience. To this aim, several models have been proposed to describe the activity of single neurons at different levels of detail. Then, single neurons are linked together to build a network, in order to reproduce complex behaviors. In the literature, different network-building rules and models have been described, targeting realistic distributions and connections of the neurons. In particular, the Granular layEr Simulator (GES) performs the granular layer network reconstruction considering biologically realistic rules to connect the neurons. Moreover, it simulates the network considering the Hodgkin–Huxley model. The work proposed in this paper adopts the network reconstruction model of GES and proposes a simulation module based on Leaky Integrate and Fire (LIF) model. This simulator targets the reproduction of the activity of large scale networks, exploiting the GPU technology to reduce the processing times. Experimental results show that a multi-GPU system reduces the simulation of a network with more than 1.8 million neurons from approximately 54 to 13 h.

Hybrid MPI and CUDA paralleled finite volume unstructured CFD simulations on a multi-GPU system

Porting unstructured Computational Fluid Dynamics (CFD) analysis of compressible flow to Graphics Processing Units (GPUs) confronts two difficulties. Firstly, non-coalescing access to the GPU’s global memory is induced by indirect data access leading to performance loss. Secondly, data exchange among multi-GPU is complex due to data communication between processes and transfer between host and device, which degrades scalability. For increasing data locality on unstructured finite volume GPU simulations for compressible flow, we perform some optimizations, including cell and face renumbering, data dependence resolving, nested loops split, and loop mode adjustment. Then, a hybrid MPI-CUDA parallel framework with packing and unpacking exchange data on GPU is established for multi-GPU computing. Finally, after optimizations, the performance of the whole application on a GPU is increased by around 50%. Simulations of ONERA M6 cases on a single GPU (Nvidia Tesla V100) can achieve an average of 13.4 speedup compared to those on 28 CPU cores (Intel Xeon Gold 6132). On the baseline of 2 GPUs, strong scaling results show a parallel efficiency of 42% on 200 GPUs, while weak scaling tests give a parallel efficiency of 82.4% up to 200 GPUs.

Computational scatter correction in near real-time with a fast Monte Carlo photon transport model for high-resolution flat-panel CT

In computed tomography (CT), scattering causes server quality degradation of the reconstructed CT images by introducing streaks and cupping artifacts which reduce the detectability of low contrast objects. Monte Carlo (MC) simulation is considered the most accurate approach for scatter estimation. However, the existing MC estimators are computationally expensive, especially for high-resolution flat-panel CT. In this paper, we propose a fast and accurate MC photon transport model which describes the physics within the 1 keV to 1 MeV range using multiple controllable key parameters. Based on this model, scatter computation for a single projection can be completed within a range of a few seconds under well-defined model parameters. Smoothing and interpolation are performed on the estimated scatter to accelerate the scatter calculation without compromising accuracy too much compared to measured near scatter-free projection images. Combining the fast scatter estimation with the filtered backprojection (FBP), scatter correction is performed effectively in an iterative manner. To evaluate the proposed MC model, we have conducted extensive experiments on the simulated data and real-world high-resolution flat-panel CT. Compared to the state-of-the-art MC simulators, the proposed MC model achieved a 15times acceleration on a single-GPU compared to the GPU implementation of the Penelope simulator (MCGPU) utilizing several acceleration techniques, and a 202 times speed-up on a multi-GPU system compared to the multi-threaded state-of-the-art EGSnrc MC simulator. Furthermore, it is shown that for high-resolution images, scatter correction with sufficient accuracy is accomplished within one to three iterations using a FBP and the proposed fast MC photon transport model.

Real-time processing pipeline for automatic streak detection in astronomical images implemented in a multi-GPU system

Abstract Detecting and tracking objects in low Earth orbit is an increasingly important task. Telescope observations contribute to its accomplishment, and telescope imagers produce a large amount of data for this task. Thus, it is convenient to use fast computer-aided processes to analyze it. Telescopes tracking at the sidereal rate usually detect these objects in their imagers as streaks, their lengths depending on the exposure time and the slant range to the object. We have developed a processing pipeline to automatically detect streaks in astronomical images in real time (i.e., faster than the images are produced) by a graphics processing unit parallel processing system. After the detection stage, streak photometric information is obtained, and object candidate identification is provided through matches with a two-line element set database. The system has been tested on a large set of images, consisting of two hours of observation time, from the Tomo-e Gozen camera of the 105 cm Schmidt telescope at Kiso Observatory in Japan. Streaks were automatically detected in approximately 0.5% of the images. The process detected streaks down to a minimum apparent magnitude of +11.3 and matched the streaks with objects from the space-track catalog in 78% of the cases. We believe that this processing pipeline can be instrumental in detecting new objects and tracking existing ones when processing speed is important, for instance, when a short handover time is required between follow-up observation stations, or when there is a large number of images to process. This study will contribute to consolidating optical observations as an effective way to control and alleviate the space debris problem.

Evolutionary background-oriented schlieren tomography with self-adaptive parameter heuristics.

For volumetric reconstruction of the refractive index field in a flow, background-oriented schlieren (BOS) imaging which measures the deflection of light rays due to refractive index variations is combined with an evolutionary tomographic algorithm for the first time, called evolutionary BOS tomography (EBOST). In this work application to reactive flows is presented. Direct non-linear ray-tracing of the reconstruction domain is used to evaluate the fitness of solution candidates during the evolutionary strategy that was implemented to run on a multi-GPU system. The use of a diversity measure and its consideration in a migration policy was tested against a simple scheme that distributes the best chromosome (solution candidate) in an island-based genetic algorithm. The extensive set of control parameters of the presented algorithm was harnessed by a self-adaptive strategy taking into account the fitness function and operator rates. Quantitative characterisation of the EBOST via numerical phantom studies, using flame simulations as ground truth data is presented. A direct comparison to a state-of-the-art BOST algorithm demonstrates similar accuracy for a turbulent swirl flame phantom reconstruction. A series of experimental applications of the EBOST on several unsteady and turbulent flames is also presented. In all cases, the instantaneous and time-averaged flame structure is revealed, proving the benefit of EBOST for volumetric flow diagnostics.

Towards scaling community detection on distributed-memory heterogeneous systems

In most real-world networks, nodes/vertices tend to be organized into tightly-knit modules known as communities or clusters such that nodes within a community are more likely to be connected or related to one another than they are to the rest of the network. Community detection in a network (graph) is aimed at finding a partitioning of the vertices into communities. The goodness of the partitioning is commonly measured using modularity. Maximizing modularity is an NP-complete problem. In 2008, Blondel et al. introduced a multi-phase, multi-iteration heuristic for modularity maximization called the Louvain method. Owing to its speed and ability to yield high quality communities, the Louvain method continues to be one of the most widely used tools for serial community detection.Distributed multi-GPU systems pose significant challenges and opportunities for efficient execution of parallel applications. Graph algorithms, in particular, have been known to be harder to parallelize on such platforms, due to irregular memory accesses, low computation to communication ratios, and load balancing problems that are especially hard to address on multi-GPU systems.In this paper, we present our ongoing work on distributed-memory implementation of Louvain method on heterogeneous systems. We build on our prior work parallelizing the Louvain method for community detection on traditional CPU-only distributed systems without GPUs. Corroborated by an extensive set of experiments on multi-GPU systems, we demonstrate competitive performance to existing distributed-memory CPU-based implementation, up to 3.2× speedup using 16 nodes of OLCF Summit relative to two nodes, and up to 19× speedup relative to the NVIDIA RAPIDS® cuGraph® implementation on a single NVIDIA V100 GPU from DGX-2 platform, while achieving high quality solutions comparable to the original Louvain method. To the best of our knowledge, this work represents the first effort for community detection on distributed multi-GPU systems. Our approach and related findings can be extended to numerous other iterative graph algorithms on multi-GPU systems.

Efficient microscopy image analysis on CPU-GPU systems with cost-aware irregular data partitioning

The analysis of high resolution whole slide tissue images is a computationally expensive task, which adversely impacts effective use of pathology imaging data in research. We propose runtime solutions to enable efficient execution of pathology image analysis applications on modern distributed memory hybrid platforms equipped with both CPUs and GPUs. Hybrid systems offer significant computation capacity, but taking advantage of this computing power is complex. An application developer may have to implement multiple versions of data processing codes targeted for different computing devices. The developer also has to tackle the challenges of efficiently distributing computational load among the nodes of a distributed memory machine and among computing devices on a node. This is particularly difficult in analysis of high resolution images because of irregular computing costs of processing different image regions. In order to address these problems, we have leveraged a high-level image processing language (Halide) and integrated it into our runtime system called Region Templates (RT). The language simplifies the application development while generating code for multiple devices, such as CPU and GPU. The integration with RT allows for efficient multiple node hybrid execution. We also developed a novel cost-aware data partitioning (CADP) strategy that considers the workload irregularity to minimize load imbalance. Our experimental evaluation shows significant performance improvements on hybrid CPU-GPU machines, as compared with using a single processor (CPU or GPU), as well as on multi-GPU systems. CADP resulted in 1.7× better performance than other workload partitioning approaches (e.g., KD-Trees) on a hybrid machine and was up to 2.24× faster in multi-node settings.

FL-MISR: fast large-scale multi-image super-resolution for computed tomography based on multi-GPU acceleration

Multi-image super-resolution (MISR) usually outperforms single-image super-resolution (SISR) under a proper inter-image alignment by explicitly exploiting the inter-image correlation. However, the large computational demand encumbers the deployment of MISR in practice. In this work, we propose a distributed optimization framework based on data parallelism for fast large-scale MISR using multi-GPU acceleration named FL-MISR. The scaled conjugate gradient (SCG) algorithm is applied to the distributed subfunctions and the local SCG variables are communicated to synchronize the convergence rate over multi-GPU systems towards a consistent convergence. Furthermore, an inner-outer border exchange scheme is performed to obviate the border effect between neighboring GPUs. The proposed FL-MISR is applied to the computed tomography (CT) system by super-resolving the projections acquired by subpixel detector shift. The SR reconstruction is performed on the fly during the CT acquisition such that no additional computation time is introduced. FL-MISR is extensively evaluated from different aspects and experimental results demonstrate that FL-MISR effectively improves the spatial resolution of CT systems in modulation transfer function (MTF) and visual perception. Comparing to a multi-core CPU implementation, FL-MISR achieves a more than 50times speedup on an off-the-shelf 4-GPU system.

Prismatic 2.0 – Simulation software for scanning and high resolution transmission electron microscopy (STEM and HRTEM)

Scanning transmission electron microscopy (STEM), where a converged electron probe is scanned over a sample's surface and an imaging, diffraction, or spectroscopic signal is measured as a function of probe position, is an extremely powerful tool for materials characterization. The widespread adoption of hardware aberration correction, direct electron detectors, and computational imaging methods have made STEM one of the most important tools for atomic-resolution materials science. Many of these imaging methods rely on accurate imaging and diffraction simulations in order to interpret experimental results. However, STEM simulations have traditionally required large calculation times, as modeling the electron scattering requires a separate simulation for each of the typically millions of probe positions. We have created the Prismatic simulation code for fast simulation of STEM experiments with support for multi-CPU and multi-GPU (graphics processing unit) systems, using both the conventional multislice and our recently-introduced PRISM method. In this paper, we introduce Prismatic version 2.0, which adds many new algorithmic improvements, an updated graphical user interface (GUI), post-processing of simulation data, and additional operating modes such as plane-wave TEM. We review various aspects of the simulation methods and codes in detail and provide various simulation examples. Prismatic 2.0 is freely available both as an open-source package that can be run using a C++ or Python command line interface, or GUI, as well within a Docker container environment.

Parallelization on Gauss Sieve Algorithm over Ideal Lattice

Cryptanalysis of lattice-based cryptography is an important field in cryptography since lattice problems are among the most robust assumptions and have been used to construct a variety of cryptographic primitives. The security estimation model for concrete parameters is one of the most important topics in lattice-based cryptography. In this research, we focus on the Gauss Sieve algorithm proposed by Micciancio and Voulgaris, a heuristic lattice sieving algorithm for the central lattice problem, shortest vector problem (SVP).We propose a technique of lifting computations in prime-cyclotomic ideals into that in cyclic ideals. Lifting makes rotations easier to compute and reduces the complexity of inner products from O(n^3) to O(n^2). We implemented the Gauss Sieve on multi-GPU systems using two layers of parallelism in our framework, and achieved up to 55 times speed of previous results of dimension 96. We were able to solve SVP on ideal lattice in dimension up to 130, which is the highest dimension SVP instance solved by sieve algorithm so far. As a result, we are able to provide a better estimate of the complexity of solving central lattice problem.