904 research outputs found
FASTCUDA: Open Source FPGA Accelerator & Hardware-Software Codesign Toolset for CUDA Kernels
No full textUsing FPGAs as hardware accelerators that communicate with a central CPU is becoming a common practice in the embedded design world but there is no standard methodology and toolset to facilitate this path yet. On the other hand, languages such as CUDA and OpenCL provide standard development environments for Graphical Processing Unit (GPU) programming. FASTCUDA is a platform that provides the necessary software toolset, hardware architecture, and design methodology to efficiently adapt the CUDA approach into a new FPGA design flow. With FASTCUDA, the CUDA kernels of a CUDA-based application are partitioned into two groups with minimal user intervention: those that are compiled and executed in parallel software, and those that are synthesized and implemented in hardware. A modern low power FPGA can provide the processing power (via numerous embedded micro-CPUs) and the logic capacity for both the software and hardware implementations of the CUDA kernels. This paper describes the system requirements and the architectural decisions behind the FASTCUDA approach
Analysis and design of massively parallel channel estimation algorithms on graphic cards
No full textThe necessity of accurate channel estimation for coherent multiuser detectors is well known. Indeed they are based on the assumption that signals are perfectly estimated, and this is never completely achieved in practice. Furthermore, practical transmitters and receivers are affected by many non-idealities like strong phase noise, and thus the task of channel estimation is all the more challenging. Another notorious issue is the high computational complexity of multiuser techniques. This project has devoted significant attention for massively parallel receiver architectures and the possibility to parallelize channel estimation algorithms. Nvidia CUDA graphic cards are especially well-suited to address problems that can be expressed as data parallel computations. This task is very challenging and ambitious, since the usage of such cards for receiver design is still at its infant stage.
This thesis describes the work carried out at German Aerospace Center (DLR) where a real-world multiuser detector is studied. The desired goals were the following: fine tuning of the already existing channel estimation algorithm; exploration of the factor graph approach in order to improve the estimation quality and to develop algorithms suitable to be parallelized; parallel implementation of the algorithms on CUDA graphic card.
All these points have been covered. Two different improvements for the already implemented phase estimator are proposed. Both are based on the same approximation of the Wiener-Levy phase model and assume the same knowledge at the receiver.
By adopting the factor graph approach, we present two existing algorithms for the phase estimation in a new parallel fashion and we show that, at the same time, they improve the estimation quality, and they are suitable to be parallelized on the board.
The performance improvement for all estimators proposed in terms of Mean Square Error are validated through several simulation campaigns carried out in different scenarios, most of them characterized by strong phase noise and low signal-to-noise ratios. Finally we present several parallel phase estimation algorithms working on CUDA graphic card and we show that, in some cases, we are in presence of a massive parallelization in which is achieved a speedup more than 200 times compared to the serial implementation. The results obtained represent a starting point for the implementation of a Parallel Iterative Receiver to be inserted in the existing multiuser detector and completely executed on CUDA graphic car
Supramolecular approaches to cryorelaxors for biological NMR studies
No full textThe work reported in this thesis concerns the development of Cryorelaxors having potential applications in very low temperature NMR studies of biological materials. Two approaches have been explored, namely calixarene host-guest complexes incorporating rotationally labile methyl groups, and endohedral fullerene complexes.The first chapter provides an overview of the background of the project and of potential approaches to the problem of low temperature relaxation agents, particularly those based on supramolecular systems. The next chapter describes the preparation, characterization, and studies of the behaviour of calixarene complexes incorporating potentially freely rotating guests. Ten calixarene complexes have been prepared and characterised by X-ray structural studies and their relaxation behaviour has been investigated using a variety of techniques including MAS and Field Cycling NMR experiments and Inelastic Neutron Scattering spectroscopy. Crystallographically disordered methyl-group environments have been identified in a number of cases and their proton spin-lattice relaxation behaviour has been shown to be strongly dependent on the thermal history of the sample. NMR studies of samples of partially deuterated calixarene complexes reveal a systematic reduction in T1 at low temperatures with higher levels of deuteration of the system studied, the p-iso-propylcalix[4]arene (2:1) p-xylene complex has been shown to have the best relaxation properties. Synthetic strategies towards calixarene and fullerenes complexes bearing active biological anchors are also discussed. The third chapter describes an investigation of the Komatsu route to the endohedral fullerene complex H2@C60 and the structural characterization of one of the key intermediates in this process. The fourth chapter give a brief overview the results obtained and the final chapter details of the experimental work undertaken. Appendices are also included containing full details of the X-ray structural studies and mathematical model
GPU acceleration for statistical gene classification
No full textThe use of Bioinformatic tools in routine clinical diagnostics is still facing a number of issues. The more complex and advanced bioinformatic tools become, the more performance is required by the computing platforms. Unfortunately, the cost of parallel computing platforms is usually prohibitive for both public and small private medical practices. This paper presents a successful experience in using the parallel processing capabilities of Graphical Processing Units (GPU) to speed up bioinformatic tasks such as statistical classification of gene expression profiles. The results show that using open source CUDA programming libraries allows to obtain a significant increase in performances and therefore to shorten the gap between advanced bioinformatic tools and real medical practic
GPU cards as a low cost solution for efficient and fast classification of high dimensional gene expression datasets
No full textThe days when bioinformatics tools will be so reliable to become a standard aid in routine clinical diagnostics are getting very close. However, it is important to remember that the more complex and advanced bioinformatics tools become, the more performances are required by the computing platforms. Unfortunately, the cost of High Performance Computing (HPC) platforms is still prohibitive for both public and private medical practices. Therefore, to promote and facilitate the use of bioinformatics tools it is important to identify low-cost parallel computing solutions. This paper presents a successful experience in using the parallel processing capabilities of Graphical Processing Units (GPU) to speed up classification of gene expression profiles. Results show that using open source CUDA programming libraries allows to obtain a significant increase in performances and therefore to shorten the gap between advanced bioinformatics tools and real medical practic
CUDA-accelerated Computational Fluid Dynamics (NVIDIA Webinar)
No full textThis webinar discusses how NVIDIA GPUs and NVIDIA CUDA can enable high-fidelity Computational Fluid Dynamics for Higher Education and Research.
Flow field computations for transient and turbulent flow problems are very compute-intensive and time-consuming. Popular existing numerical techniques often compromise on the underlying physics or require a massive amount of computational resources. Accurate high-fidelity CFD simulations on locally available hardware hence are highly appreciated by academia and industry.
After a brief review of the theoretical basics and implementation details of a CUDA-accelerated CFD solver, results from several international research projects are presented. They all demonstrate that GPU computing can be a game changer for state-of-the art research projects in many relevant areas of Computational Fluid Dynamics.</p
Parallelising wavefront applications on general-purpose GPU devices
No full textPipelined wavefront applications form a large portion of the high performance scientific computing workloads at supercomputing centres. This paper investigates the viability of graphics processing units (GPUs) for the acceleration of these codes, using NVIDIA's Compute Unified Device Architecture (CUDA). We identify the optimisations suitable for this new architecture and quantify the characteristics of those wavefront codes that are likely to experience speedups
Acceleration computing process in wavelength scanning interferometry
No full textThe optical interferometry has been widely explored for surface measurement due to the advantages of non-contact and high accuracy interrogation. Eventually, some interferometers are used to measure both rough and smooth surfaces such as white light interferometry and wavelength scanning interferometry (WSI). The WSI can be used to measure large discontinuous surface profiles without the phase ambiguity problems. However, the WSI usually needs to capture hundreds of interferograms at different wavelength in order to evaluate the surface finish for a sample. The evaluating process for this large amount of data needs long processing time if CPUs traditional programming is used. This paper presents a parallel programming model to achieve the data parallelism for accelerating the computing analysis of the captured data. This parallel programming is based on CUDATM C program structure that developed by NVIDIA. Additionally, this paper explains the mathematical algorithm that has been used for evaluating the surface profiles. The computing time and accuracy obtained from CUDA program, using GeForce GTX 280 graphics processing unit (GPU), were compared to those obtained from sequential execution Matlab program, using Intel® Core™2 Duo CPU. The results of measuring a step height sample shows that the parallel programming capability of the GPU can highly accelerate the floating point calculation throughput compared to multicore CPU
- …
