Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “GPUs”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

Efficient Scheduling of Recursive Control Flow on GPUs

Graphics processing units (GPUs) have rapidly emerged as a very significant player in high performance computing. Single instruction multiple thread (SIMT) pipelines are typically used in GPUs to exploit parallelism and maximize performance. Although support for unstructured control flow has been included in GPUs, efficiently managing thread divergence for arbitrary parallel programs remains a critical challenge. In this paper, we focus on the problem of supporting recursion in modern GPUs. We design and comparatively evaluate various algorithms to manage thread divergence encountered in recursive programs. The results improve upon traditional post-dominator based reconvergence mechanisms designed to handle thread divergence due to control flow within a procedure.

GPUs↗

High performance sparse multifrontal solvers on modern GPUs

Here, we have ported the numerical factorization and triangular solve phases of the sparse direct solver STRUMPACK to GPU. STRUMPACK implements sparse LU factorization using the multifrontal algorithm, which performs most of its operations in dense linear algebra operations on so-called frontal matrices of various sizes. Our GPU implementation off-loads these dense linear algebra operations, as well as the sparse scatter–gather operations between frontal matrices. For the larger frontal matrices, our GPU implementation relies on vendor libraries such as cuBLAS and cuSOLVER for NVIDIA GPUs and rocBLAS and rocSOLVER for AMD GPUs. For the smaller frontal matrices we developed custom CUDA and HIP kernels to reduce kernel launch overhead. Overall, high performance is achieved by identifying submatrix factorizations corresponding to sub-trees of the multifrontal assembly tree which fit entirely in GPU memory. The multi-GPU setting uses SLATE (Software for Linear Algebra Targeting Exascale) as a modern GPU-aware replacement for ScaLAPACK. On 4 nodes of SUMMIT the code runs ~10X faster when using all 24 V100 GPUs compared to when it only uses the 168 POWER9 cores. On 8 SUMMIT nodes, using 48 V100 GPUs, the sparse solver reaches over 50TFlop/s. Compared to SuperLU, on a single V100, for a set of 17 matrices our implementation is faster for all but one matrix, and is on average 5X (median 4X) faster

97 MATHEMATICS AND COMPUTING↗

Radiative Heat Transfer Calculation on 16384 GPUs Using a Reverse Monte Carlo Ray Tracing Approach with Adaptive Mesh Refinement

Modeling thermal radiation is computationally challenging in parallel due to its all-to-all physical and resulting computational connectivity, and is also the dominant mode of heat transfer in practical applications such as next-generation clean coal boilers, being modeled by the Uintah framework. However, a direct all-to-all treatment of radiation is prohibitively expensive on large computers systems whether homogeneous or heterogeneous. DOE Titan and the planned DOE Summit and Sierra machines are examples of current and emerging GPU-based heterogeneous systems where the increased processing capability of GPUs over CPUs exacerbates this problem. These systems require that computational frameworks like Uintah leverage an arbitrary number of on-node GPUs, while simultaneously utilizing thousands of GPUs within a single simulation. We show that radiative heat transfer problems can be made to scale within Uintah on heterogeneous systems through a combination of reverse Monte Carlo ray tracing (RMCRT) techniques combined with AMR, to reduce the amount of global communication. In particular, significant Uintah infrastructure changes, including a novel lock and contention-free, thread-scalable data structure for managing MPI communication requests and improved memory allocation strategies were necessary to achieve excellent strong scaling results to 16384 GPUs on Titan.

Computer Science↗

Accelerating Binarized Neural Networks via Bit-Tensor-Cores in Turing GPUs

Despite foreseeing tremendous speedups over conventional deep neural networks, the performance advantage of binarized neural networks (BNNs) has merely been showcased on general-purpose processors such as CPUs and GPUs. In fact, due to being unable to leverage bit-level-parallelism with a word-based architecture, GPUs have been criticized for extremely low utilization (1%) when executing BNNs. Consequently, the latest tensorcores in NVIDIA Turing GPUs start to experimentally support bit computation. In this work, we look into this brand new bit computation capability and characterize its unique features. We show that the stride of memory access can significantly affect performance delivery and a data-format co-design is highly desired to support the tensorcores for achieving superior performance than existing software solutions without tensorcores. We realize the tensorcore-accelerated BNN design, particularly the major functions for fully-connect and convolution layers — bit matrix multiplication and bit convolution. Evaluations on two NVIDIA Turing GPUs show that, with ResNet-18, our BTC-BNN design can process ImageNet at a rate of 5.6K images per second, 77% faster than state-of-the-art. Our BNN approach is released on https://github.com/pnnl/TCBNN.

Li, Ang↗

Fast DRR generation for 2D to 3D registration on GPUs

Purpose: The generation of digitally reconstructed radiographs (DRRs) is the most time consuming step on the CPU in intensity based two-dimensional x-ray to three-dimensional (CT or 3D rotational x-ray) medical image registration, which has application in several image guided interventions. This work presents optimized DRR rendering on graphical processor units (GPUs) and compares performance achievable on four commercially available devices. Methods: A ray-cast based DRR rendering was implemented for a 512 Multiplication-Sign 512 Multiplication-Sign 72 CT volume. The block size parameter was optimized for four different GPUs for a region of interest (ROI) of 400 Multiplication-Sign 225 pixels with different sampling ratios (1.1%-9.1% and 100%). Performance was statistically evaluated and compared for the four GPUs. The method and the block size dependence were validated on the latest GPU for several parameter settings with a public gold standard dataset (512 Multiplication-Sign 512 Multiplication-Sign 825 CT) for registration purposes. Results: Depending on the GPU, the full ROI is rendered in 2.7-5.2 ms. If sampling ratio of 1.1%-9.1% is applied, execution time is in the range of 0.3-7.3 ms. On all GPUs, the mean of the execution time increased linearly with respect to the number of pixels if sampling was used. Conclusions: The presented results outperform other results from the literature. This indicates that automatic 2D to 3D registration, which typically requires a couple of hundred DRR renderings to converge, can be performed quasi on-line, in less than a second or depending on the application and hardware in less than a couple of seconds. Accordingly, a whole new field of applications is opened for image guided interventions, where the registration is continuously performed to match the real-time x-ray.

07 ISOTOPE AND RADIATION SOURCES↗

Mojo: MLIR-based Performance-Portable HPC Science Kernels on GPUs for the Python Ecosystem

We explore the performance and portability of the novel Mojo language for scientific computing workloads on GPUs. As the first language based on the LLVM’s Multi-Level Intermediate Representation (MLIR) compiler infrastructure, Mojo aims to close performance and productivity gaps by combining Python’s interoperability and CUDA-like syntax for compile-time portable GPU programming. We target four scientific workloads: a seven-point stencil (memory-bound), BabelStream (memory-bound), miniBUDE (compute-bound), and Hartree–Fock (compute-bound with atomic operations); and compare their performance against vendor baselines on NVIDIA H100 and AMD MI300A GPUs. We show that Mojo’s performance is competitive with CUDA and HIP for memory-bound kernels, whereas gaps exist on AMD GPUs for atomic operations and for fast-math compute-bound kernels on both AMD and NVIDIA GPUs. Although the learning curve and programming requirements are still fairly low-level, Mojo can close significant gaps in the fragmented Python ecosystem in the convergence of scientific computing and AI.

Godoy, William [ORNL] (ORCID:0000000225905178)↗

Portable Software Environment for Ultrahigh-Resolution ELM Development on GPUs

This paper presents our endeavors in developing the large-scale, ultra-high-resolution E3SM Land Model (uELM), specifically designed for exascale computers furnished with accelerators such as Nvidia GPUs. The uELM is a sophisticated code that substantially relies on High-Performance Computing (HPC) environments, necessitating particular machine and software configurations. To facilitate community-based uELM developments employing GPUs, we have created a portable, standalone software environment preconfigured with uELM input datasets, simulation cases, and source code. This environment, utilizing Docker, encompasses all essential code, libraries, and system software for uELM development on GPUs. It also features a functional unit test framework and an offline model testbed for comprehensive numerical experiments. From a technical perspective, the paper discusses GPU-ready container generations, uELM code management, and input data distribution across computational platforms. Lastly, the paper demonstrates the use of environment for functional unit testing, end-to-end simulation on CPUs and GPUs, and collaborative code development.

E3SM Land Model↗

Implementation of a Particle Accelerator Beam Dynamics Code on Multi-Node GPUs

Particle accelerators play an important role in a wide range of scientific discoveries and industrial applications. The self-consistent multi-particle simulation based on the particle-in-cell (PIC) method has been used to study charged particle beam dynamics inside those accelerators. However, the PIC simulation is time-consuming and needs to use modern parallel computers for high-resolution applications. In this paper, we implemented a parallel beam dynamics PIC code on multi-node hybrid architecture computers with multiple Graphics Processing Units (GPUs). We used two methods to parallelize the PIC code on multiple GPUs and observed that the replication method is a better choice for moderate problem size and current computer hardware while the domain decomposition method might be a better choice for large problem size and more advanced computer hardware that allows direct communications among multiple GPUs. Using the multi-node hybrid architectures at Oak Ridge Leadership Computing Facility (OLCF), the optimized GPU PIC code achieves a reasonable parallel performance and scales up to 64 GPUs with 16 million particles.

43 PARTICLE ACCELERATORS↗

Performance study of GPUs in real-time trigger applications for HEP experiments

Graphical Processing Units (GPUs) have evolved into highly parallel, multi-threaded, multicore powerful processors with high memory bandwidth. GPUs are used in a variety of intensive computing applications. The combination of highly parallel architecture and high memory bandwidth makes GPUs a potentially promising technology for effective real-time processing for High Energy Physics (HEP) experiments. However, not much is known of their performance in real-time applications that require low latency, such as the trigger for HEP experiments. We describe an R and D project with the goal to study the performance of GPU technology for possible low latency applications, performing basic operations as well as some more advanced HEP lower-level trigger algorithms (such as fast tracking or jet finding). We present some preliminary results on timing measurements, comparing the performance of a CPU versus a GPU with NVIDIA's CUDA general-purpose parallel computing architecture, carried out at CDF's Level-2 trigger test stand. These studies will provide performance benchmarks for future studies to investigate the potential and limitations of GPUs for real-time applications in HEP experiments.

43 PARTICLE ACCELERATORS↗

Data Acquisition with GPUs: The DAQ for the Muon $g$-$2$ Experiment at Fermilab

Graphical Processing Units (GPUs) have recently become a valuable computing tool for the acquisition of data at high rates and for a relatively low cost. The devices work by parallelizing the code into thousands of threads, each executing a simple process, such as identifying pulses from a waveform digitizer. The CUDA programming library can be used to effectively write code to parallelize such tasks on Nvidia GPUs, providing a significant upgrade in performance over CPU based acquisition systems. The muon $g$-$2$ experiment at Fermilab is heavily relying on GPUs to process its data. The data acquisition system for this experiment must have the ability to create deadtime-free records from 700 $\mu$s muon spills at a raw data rate 18 GB per second. Data will be collected using 1296 channels of $\mu$TCA-based 800 MSPS, 12 bit waveform digitizers and processed in a layered array of networked commodity processors with 24 GPUs working in parallel to perform a fast recording of the muon decays during the spill. The described data acquisition system is currently being constructed, and will be fully operational before the start of the experiment in 2017.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Enhancing Monte Carlo proxy applications on GPUs

In Monte Carlo neutron transport simulations, a computational routine commonly known as the ”cross-section lookup” has been identified as being the most computationally expensive part of these applications. A tool which is commonly used as a proxy application for these routines, named ”XSBench”, was created to simulate popular algorithms used in these routines on CPUs. Currently, however, as GPU-based HPC resources have become more widely available, there has been significant interest and efforts invested in moving these traditionally CPU-based simulations to GPUs. Unfortunately, the algorithms commonly used in the cross-section lookup routine were originally devised and developed for CPU-based platforms, and have seen limited study on GPUs to date. Additionally, platforms such as XSBench implement approximations which may have a negligible effect on CPUs, but may be quite impactful to performance on GPUs given the more resource-limited nature of the latter. As a result, we have created VEXS, a new tool for modeling the cross-section lookup routine which removes or at least reduces the approximations made by XSBench in order to provide a more realistic prediction of algorithm performance on GPUs. In this paper, we detail our efforts to remove and reduce these approximations, show the resulting improvement in performance prediction in comparison to a reference production code, Shift, and provide some basic profiling analysis of the resulting application.

Shriver, Forrest↗

m-CUBES An efficient and portable implementation of multi-dimensional integration for gpus

The task of multi-dimensional numerical integration is frequently encountered in physics and other scientific fields, e.g., in modeling the effects of systematic uncertainties in physical systems and in Bayesian parameter estimation. Multi-dimensional integration is often time-prohibitive on CPUs. Efficient implementation on many-core architectures is challenging as the workload across the integration space cannot be predicted a priori. We propose m-Cubes, a novel implementation of the well-known Vegas algorithm for execution on GPUs. Vegas transforms integration variables followed by calculation of a Monte Carlo integral estimate using adaptive partitioning of the resulting space. m-Cubes improves performance on GPUs by maintaining relatively uniform workload across the processors. As a result, our optimized Cuda implementation for Nvidia GPUs outperforms parallelization approaches proposed in past literature. We further demonstrate the efficiency of m-Cubes by evaluating a six-dimensional integral from a cosmology application, achieving significant speedup and greater precision than the CUBA library's CPU implementation of VEGAS. We also evaluate m-Cubes on a standard integrand test suite. m-Cubes outperforms the serial implementations of the Cuba and GSL libraries by orders of magnitude speedup while maintaining comparable accuracy. Our approach yields a speedup of at least 10 when compared against publicly available Monte Carlo based GPU implementations. In summary, m-Cubes can solve integrals that are prohibitively expensive using standard libraries and custom implementations. A modern C++ interface header-only implementation makes m-Cubes portable, allowing its utilization in complicated pipelines with easy to define stateful integrals. Compatibility with non-Nvidia GPUs is achieved with our initial implementation of m-Cubes using the Kokkos framework.

Sakiotis, Ioannis↗

A massively parallel method of characteristic neutral particle transport code for GPUs

Over the past 20 years, parallel computing has enabled computers to grow ever larger and more powerful while scientific applications have advanced in sophistication and resolution. This trend is being challenged, however, as the power consumption for conventional parallel computing architectures has risen to unsustainable levels and memory limitations have come to dominate compute performance. Heterogeneous computing platforms, such as Graphics Processing Units (GPUs), are an increasingly popular paradigm for solving these issues. This paper explores the applicability of GPUs for deterministic neutron transport. A 2D method of characteristics (MOC) code - OpenMOC - has been developed with solvers for both shared memory multi-core platforms as well as GPUs. The multi-threading and memory locality methodologies for the GPU solver are presented. Performance results for the 2D C5G7 benchmark demonstrate 25-35 x speedup for MOC on the GPU. The lessons learned from this case study will provide the basis for further exploration of MOC on GPUs as well as design decisions for hardware vendors exploring technologies for the next generation of machines for scientific computing. (authors)

97 MATHEMATICS AND COMPUTING↗

Optimal size of the block in block GMRES on GPUs: computational model and experiments

The block version of GMRES (BGMRES) is most advantageous over the single right hand side (RHS) counterpart when the cost of communication is high while the cost of floating point operations is not. This is the particular case on modern graphics processing units (GPUs), while it is generally not the case on traditional central processing units (CPUs). Here, in this paper, experiments on both GPUs and CPUs are shown that compare the performance of BGMRES against GMRES as the number of RHS increases, with a particular focus on GPU performance. The experiments indicate that there are many cases in which BGMRES is slower than GMRES on CPUs, but faster on GPUs. Furthermore, when varying the number of RHS on the GPU, there is an optimal number of RHS where BGMRES is clearly most advantageous over GMRES. A computational model for the GPU is developed using hardware specific parameters, providing insight towards how the qualitative behavior of BGMRES changes as the number of RHS increase, and this model also helps explain the phenomena observed in the experiments.

97 MATHEMATICS AND COMPUTING↗

Mesoscopic Modeling and Rapid Simulation of Incremental Changes in Epidemic Scenarios on GPUs

In simulation-based studies and analyses of epidemics, a major challenge lies in resolving the conflict between fidelity of models and the speed of their simulation. Another related challenge arises in dealing with the large number of what–if scenarios that need to be explored. Here, we describe new computational methods that together provide an approach to dealing with both challenges. A mesoscopic modeling approach is described that strikes a middle ground between macroscopic models based on coupled differential equations and microscopic models built on fine-grained behaviors at the individual entity level. The mesoscopic approach offers the ability to incorporate complex compositions of multiple layers of dynamics even while retaining the potential for aggregate behaviors at varying levels. It also is an excellent match to the accelerator-based architectures of modern computing platforms in which graphical processing units (GPUs) can be exploited for fast simulation via the parallel execution mode of single instruction multiple thread (SIMT). The challenge of simulating a large number of scenarios is addressed via a method of sharing model state and computation across a tree of what–if scenarios that are localized, incremental changes to a large base simulation. A combination of the mesoscopic modeling approach and the incremental what–if scenario tree evaluation has been implemented in the software on modern GPUs. Synthetic simulation scenarios are presented to demonstrate the computational characteristics of our approach. Results from the experiments with large population data, including USA, UK, and India, illustrate the modeling methodology and computational performance on thousands of synthetically generated what–if scenarios. Execution of our implementation scaled to 8192 GPUs of supercomputing platforms demonstrates the ability to rapidly evaluate what–if scenarios several orders of magnitude faster than the conventional methods.

97 MATHEMATICS AND COMPUTING↗

Symplectic multi-particle tracking on GPUs

A symplectic multi-particle tracking model is implemented on the Graphic Processing Units (GPUs) using the Compute Unified Device Architecture (CUDA) language. The symplectic tracking model can preserve phase space structure and reduce non-physical effects in long term simulation, which is important for beam property evaluation in particle accelerators. Though this model is computationally expensive, it is very suitable for parallelization and can be accelerated significantly by using GPUs. In this paper, we optimized the implementation of the symplectic tracking model on both single GPU and multiple GPUs. Using a single GPU processor, the code achieves a factor of 2-10 speedup for a range of problem sizes compared with the time on a single state-of-the-art Central Processing Unit (CPU) node with similar power consumption and semiconductor technology. It also shows good scalability on a multi-GPU cluster at Oak Ridge Leadership Computing Facility. In an application to beam dynamics simulation, the GPU implementation helps save more than a factor of two total computing time in comparison to the CPU implementation.

43 PARTICLE ACCELERATORS↗

Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience

We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GenASiS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ~ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.

97 MATHEMATICS AND COMPUTING↗

Porting fragmentation methods to GPUs using an OpenMP API: Offloading the resolution-of-the-identity second-order Møller–Plesset perturbation method

Here, using an OpenMP Application Programming Interface, the resolution-of-the-identity second-order Møller–Plesset perturbation (RI-MP2) method has been off-loaded onto graphical processing units (GPUs), both as a standalone method in the GAMESS electronic structure program and as an electron correlation energy component in the effective fragment molecular orbital (EFMO) framework. First, a new scheme has been proposed to maximize data digestion on GPUs that subsequently linearizes data transfer from central processing units (CPUs) to GPUs. Second, the GAMESS Fortran code has been interfaced with GPU numerical libraries (e.g., NVIDIA cuBLAS and cuSOLVER) for efficient matrix operations (e.g., matrix multiplication, matrix decomposition, and matrix inversion). The standalone GPU RI-MP2 code shows an increasing speedup of up to 7.5× using one NVIDIA V100 GPU with one IBM 42-core P9 CPU for calculations on fullerenes of increasing size from 40 to 260 carbon atoms using the 6-31G(d)/cc-pVDZ-RI basis sets. A single Summit node with six V100s can compute the RI-MP2 correlation energy of a cluster of 175 water molecules using the correlation consistent basis sets cc-pVDZ/cc-pVDZ-RI containing 4375 atomic orbitals and 14 700 auxiliary basis functions in ~0.85 h. In the EFMO framework, the GPU RI-MP2 component shows near linear scaling for a large number of V100s when computing the energy of an 1800-atom mesoporous silica nanoparticle in a bath of 4000 water molecules. The parallel efficiencies of the GPU RI-MP2 component with 2304 and 4608 V100s are 98.0% and 96.1%, respectively.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗