Search NASA⌕ Search

Engineering topics

Finkel, Hal

Publications and source records attributed to Finkel, Hal.

At least 19 records

Autotuning PolyBench benchmarks with LLVM Clang/Polly loop optimization pragmas using Bayesian optimization

Here, we develop a ytopt autotuning framework that leverages Bayesian optimization to explore the parameter space search and compare four different supervised learning methods within Bayesian optimization and evaluate their effectiveness. We select six of the most complex PolyBench benchmarks and apply the newly developed LLVM Clang/Polly loop optimization pragmas to the benchmarks to optimize them. We then use the autotuning framework to optimize the pragma parameters to improve their performance. The experimental results show that our autotuning approach outperforms the other compiling methods to provide the smallest execution time for the benchmarks syr2k, 3mm, heat-3d, lu, and covariance with two large datasets in 200 code evaluations for effectively searching the parameter spaces with up to 170,368 different configurations. We find that the Floyd-Warshall benchmark did not benefit from autotuning. To cope with this issue, we provide some compiler option solutions to improve the performance. Then we present loop autotuning without a user's knowledge using a simple mctree autotuning framework to further improve the performance of the Floyd-Warshall benchmark. We also extend the ytopt autotuning framework to tune a deep learning application.

79 ASTRONOMY AND ASTROPHYSICS↗

OpenMP application experiences: Porting to accelerated nodes

As recent enhancements to the OpenMP specification become available in its implementations, there is a need to share the results of experimentation in order to better understand the OpenMP implementation’s behavior in practice, to identify pitfalls, and to learn how the implementations can be effectively deployed in scientific codes. We report on experiences gained and practices adopted when using OpenMP to port a variety of ECP applications, mini-apps and libraries based on different computational motifs to accelerator-based leadership-class high-performance supercomputer systems at the United States Department of Energy. Additionally, we identify important challenges and open problems related to the deployment of OpenMP. Through our report of experiences, we find that OpenMP implementations are successful on current supercomputing platforms and that OpenMP is a promising programming model to use for applications to be run on emerging and future platforms with accelerated nodes.

97 MATHEMATICS AND COMPUTING↗

Non-Proprietary Companion to the Q1 CY2021 PathForward Final Assessment (WBS 2.4.1, Milestone PM-HI-1040)

The PathForward element of the Exascale Computing Project (ECP) Hardware and Integration (HI) focus area has met its goals of preparing the US industry for exascale system procurements and generally improving US competitiveness in the worldwide computing market. This report is a companion to the deliverable for the ECP milestone PM-HI-1040, Assess PathForward Impact Against Exascale Hardware Challenges, which gives a summary of the final status of each PathForward project, describes progress achieved against PathForward contract milestones, and includes a final assessment of each vendor’s progress on key exascale challenges. This companion document details the results of the PathForward research to the extent possible without disclosing proprietary information and the impact on products and US exascale systems. In addition, it captures lessons learned in order to inform future projects in general and in high-performance computing in particular.

97 MATHEMATICS AND COMPUTING↗

The Last Journey. I. An Extreme-scale Simulation on the Mira Supercomputer

The Last Journey is a large-volume, gravity-only, cosmological N-body simulation evolving more than 1.24 trillion particles in a periodic box with a side length of 5.025 Gpc. It was implemented using the HACC simulation and analysis framework on the BG/Q system Mira. The cosmological parameters are chosen to be consistent with the results from the Planck satellite. A range of analysis tools have been run in situ to enable a diverse set of science projects and, at the same time, keep the resulting data amount manageable. Analysis outputs have been generated starting at redshift z ~ 10 to allow for construction of synthetic galaxy catalogs using a semianalytic modeling approach in postprocessing. As part of our in situ analysis pipeline, we employ a new method for tracking halo substructures, introducing the concept of subhalo cores. The production of multiwavelength synthetic sky maps is facilitated by generating particle light cones in situ, also beginning at z ~ 10. We provide an overview of the simulation setup and generated data products; a first set of analysis results is presented. A subset of the data is publicly available.

79 ASTRONOMY AND ASTROPHYSICS↗

Proxy App Suite Release (FY2020)

Version 4.0 of the ECP Proxy App Suite is practically unchanged from the previous release. The current set of proxies has proven useful for many aspects of benchmarking and co-design and we see little reason to alter the suite. Although there have been few changes to the ECP suite, the team has been hard at work in other areas. In the area of Machine Learning (ML) we have now created a separate proxy suite dedicated to this scientific applications of ML.

97 MATHEMATICS AND COMPUTING↗

FY20 Proxy App Suite Release: Report for ECP Proxy App Project Milestone ADCD-504-10

Version 4.0 of the ECP Proxy App Suite is practically unchanged from the previous release. The current set of proxies has proven useful for many aspects of benchmarking and co-design and we see little reason to alter the suite. Although there have been few changes to the ECP suite, the team has been hard at work in other areas. In the area of Machine Learning (ML) we have now created a separate proxy suite dedicated to this scientific applications of ML. The suite includes: miniGAN (Generative Adversarial Networks), miniRL (Reinforcement Learning), CRADL (inline inference), Cosmoflow-Benchmark (Convolutional Neural Network), and MLPerf-DeepCam (Climate Segmentation Benchmark). Section 3 contains more information about these proxies as well as the principles that are guiding the development of the suite. We have surveyed available proxies for several application domains including Computational Fluid Dynamics, Quantum Chemistry, Quantum Computing Simulation, Molecular Dynamics, Monte Carlo Transport, and Density Functional Theory to identify gaps in proxy coverage. Several new proxy apps are either already available or will be released soon to fill these gaps. Section 4 provides full details. Finally, section 5 reports on our continued collaboration with the ECP Continuous Integration (CI) effort to use proxy apps to help identify problems and roadblocks to cross-lab CI. We have also assisted the El Capitan Center of Excellence (COE) to stand up a CI system that will be used to test releases of HPE and AMD software stacks. We hope that the COE effort can serve as a model for ECP by showing how the Proxy App Team can work with AD and ST teams identify critical features, kernels, patterns, etc. and incorporate them into a CI system that will help ensure that Frontier and Aurora will provide those needed capabilities.

97 MATHEMATICS AND COMPUTING↗

The Mira-Titan Universe. III. Emulation of the Halo Mass Function

We construct an emulator for the halo mass function over group and cluster mass scales for a range of cosmologies, including the effects of dynamical dark energy and massive neutrinos. The emulator is based on the recently completed Mira-Titan Universe suite of cosmological N-body simulations. The main set of simulations spans 111 cosmological models with 2.1 Gpc boxes. We extract halo catalogs in the redshift range z = [0.0, 2.0] and for masses M-200c >= 10(13)M(circle dot)/h. The emulator covers an eight-dimensional hypercube spanned by {Omega(m)h(2), Omega(b)h(2), Omega(nu)h(2), sigma(8), h, n(s), w(0), w(a)}; spatial flatness is assumed. We obtain smooth halo mass functions by fitting piecewise second-order polynomials to the halo catalogs and employ Gaussian process regression to construct the emulator while keeping track of the statistical noise in the input halo catalogs and uncertainties in the regression process. For redshifts z less than or similar to 1, the typical emulator precision is better than 2% for 10(13)-10(14)M(circle dot)/h and <10<^> M similar or equal to 101(circle dot)(54)/h. For comparison, fitting functions using the traditional universal form for the halo mass function can be biased at up to 30% at M similar or equal to 10(144)M(circle dot)/h for z = 0. Our emulator is publicly available at https://github.com/SebastianBocquet/MiraTitanHMFemulator.

cosmology: theory↗

Population Count on Intel® CPU, GPU, and FPGA

Population count is a primitive used in many applications. Commodity processors have dedicated instructions for achieving high-performance population count. Motivated by the productivity of high-level synthesis and the importance of population count, in this paper we investigated the OpenCL implementations of population count algorithms, and evaluated their performance and resource utilizations on an FPGA. Based on the results, we select the most efficient implementation. Then we derived a reduction pattern from a representative application of population count. We parallelized the reduction with atomic functions, and optimized it with vectorized memory accesses, tree reduction, and compute-unit duplication. We evaluated the performance of the reduction kernel on an Intel (R) Xeon (R) CPU and an Intel (R) IrisT Pro integrated GPU, and an FPGA card that features an Intel (R) Arria (R) 10 FPGA. When DRAM memory bandwidth is comparable on the three computing platforms, the FPGA can achieve the highest kernel performance for large workload. On the other hand, we described performance bottlenecks on the FPGA. To make FPGAs more competitive in raw performance compared to high-performant CPU and GPU platforms, it is important to increase external memory bandwidth, minimize data movement between a host and a device, and reduce OpenCL runtime overhead on an FPGA.

heterogeneous computing↗

Analyzing Deep Learning Model Inferences for Image Classification using OpenVINO

It may be desirable to execute deep learning model inferences on an integrated GPU at the edge. While such GPUs are much less powerful than discrete GPUs, it is able to deliver higher floating-point operations per second than a CPU located on the same die. For edge devices, the benefit of moving to lower precision with minimal loss of accuracy to obtain higher performance is also attractive. Hence, we chose 14 deep learning models for image classification to evaluate their inference performance with the OpenVINO toolkit. Then, we analyzed the implementation of the fastest inference model of all the models. The experimental results are promising. Compared to the performance of full-precision (FP32) models, the speedup of the 8-bit (INT8) quantization ranges from 1.02 to 1.56 on an Intel (R) Xeon (R) 4-core CPU, and the speedup of the FP16 models ranges from 1.1 to 2 on an Intel (R) Iris (TM) Pro GPU. For the FP32 models, the GPU is on average 1.5X faster than the CPU.

CPU↗

A Case Study on the HACCmk Routine in SYCL on Integrated Graphics

As opposed to the Open Computing Language (OpenCL) programming model in which host and device codes are generally written in different languages, the SYCL programming model can combine host and device codes for an application in a type-safe way to improve development productivity. In this paper, we chose the HACCmk routine, a representative compute-bound kernel, as a case study on the performance of the SYCL programming model targeting a heterogeneous computing device. More specifically, we introduced the SYCL programming model, presented the OpenCL and SYCL implementations of the routine, and compared the performance of the two implementations using the offline and online compilation on Intel (R) Iris (TM) Pro integrated GPUs. We found that the overhead of online compilation may become significant compared to the execution time of a kernel. Compared to the performance of OpenCL implementations, the SYCL implementation can maintain the performance using the offline compilation. The number of execution units in a GPU are critical to improving the raw performance of a compute-bound kernel.

Compilation modes↗

CosmoDC2: A Synthetic Sky Catalog for Dark Energy Science with LSST

This paper introduces cosmoDC2, a large synthetic galaxy catalog designed to support precision dark energy science with the Large Synoptic Survey Telescope (LSST). CosmoDC2 is the starting point for the second data challenge (DC2) carried out by the LSST Dark Energy Science Collaboration (LSST DESC). The catalog is based on a trillion-particle, (4.225 Gpc) 3 box cosmological N-body simulation, the Outer Rim run. It covers 440 deg 2 of sky area to a redshift of z = 3 and matches expected number densities from contemporary surveys to a magnitude depth of 28 in the r band. Each galaxy is characterized by a multitude of galaxy properties including stellar mass, morphology, spectral energy distributions, broadband filter magnitudes, host halo information, and weak lensing shear. The size and complexity of cosmoDC2 requires an efficient catalog generation methodology; our approach is based on a new hybrid technique that combines data-based empirical approaches with semianalytic galaxy modeling. A wide range of observation-based validation tests has been implemented to ensure that cosmoDC2 enables the science goals of the planned LSST DESC DC2 analyses. This paper also represents the official release of the cosmoDC2 data set, including an efficient reader that facilitates interaction with the data.

79 ASTRONOMY AND ASTROPHYSICS↗

The Outer Rim Simulation: A Path to Many-core Supercomputers

Herein we describe the Outer Rim cosmological simulation, one of the largest high-resolution N-body simulations performed to date, aimed at promoting science to be carried out with large-scale structure surveys. The simulation covers a volume of (4.225 Gpc) 3 and evolves more than one trillion particles. It was executed on Mira, a BlueGene/Q system at the Argonne Leadership Computing Facility. We discuss some of the computational challenges posed by a system such as Mira, a many-core supercomputer, and how the simulation code, Hardware/Hybrid Accelerated Cosmology Code, has been designed to overcome these challenges. We have carried out a large range of analyses on the simulation data and we report on the results as well as the data products that have been generated. The full data set generated by the simulation totals more than 5 PB of data, making curation and handling of the data a large challenge in itself. The simulation results have been used to generate synthetic catalogs for large-scale structure surveys, including Dark Energy Spectroscopic Instrument and eBOSS, as well as experiments on the cosmic microwave background. A detailed catalog for the data challenges of the Large Synoptic Survey Telescope Dark Energy Science Collaboration has been created as well. We publicly release some of the Outer Rim halo catalogs, downsampled particle information, and lightcone data.

79 ASTRONOMY AND ASTROPHYSICS↗

Quantitative Performance Assessment of Proxy Apps and Parents (Report for ECP Proxy App Project Milestone AD-CD-PA-504-5)

Proxy applications are small, simplified codes that allow application developers to share important features of large applications without forcing collaborators to assimilate large and complex code bases. Proxies can also be though of as models for larger codes. Because proxies omit features and capabilities of their parent applications (otherwise they wouldn’t be small), the question naturally arises whether any given proxy is actually a faithful model for its parent application for some particular modeling purpose. This report sets out to answer that question for some of the proxies in the ECP proxy app collection, with particular attention to performance. It also completes the AD-CD-PA-504 Milestone: "We will assess the fidelity of 8 proxy applications with respect to their respective parent applications using the methodology previously developed."

97 MATHEMATICS AND COMPUTING↗

The Borg Cube Simulation: Cosmological Hydrodynamics with CRK-SPH

A challenging requirement posed by next-generation observations is a firm theoretical grasp of the impact of baryons on structure formation. Cosmological hydrodynamic simulations modeling gas physics are vital in this regard. A high degree of modeling flexibility exists in this space, making it important to explore a range of methods in order to gauge the accuracy of simulation predictions. We present results from the first cosmological simulation using Conservative Reproducing Kernel Smoothed Particle Hydrodynamics (CRK-SPH). We employ two simulations: one evolved purely under gravity, and the other with nonradiative hydrodynamics. Each contains 2 × 2304 3 cold dark matter plus baryon particles in an 800 h -1 Mpc box. We compare statistics to previous nonradiative simulations including power spectra, mass functions, baryon fractions, and concentration. We find self-similar radial profiles of gas temperature, entropy, and pressure and show that a simple analytic model recovers these results to better than 40% over two orders of magnitude in mass. We quantify the level of nonthermal pressure support in halos and demonstrate that hydrostatic mass estimates are biased low by 24% (10%) for halos of mass 10 15 (10 13 ) h -1 M⊙. Here, we compute angular power spectra for the thermal and kinematic Sunyaev–Zel’dovich effects and find good agreement with the low-ℓ Planck measurements. Finally, artificial scattering between particles of unequal mass is shown to have a large impact on the gravity-only run, and we highlight the importance of better understanding this issue in hydrodynamic applications. This is the first in a simulation campaign using CRK-SPH, with future work including subresolution gas treatments.

79 ASTRONOMY AND ASTROPHYSICS↗

Exploring Integer Sum Reduction using Atomics on Intel CPU

Atomic functions are useful in updating a shared variable by multiple threads, barrier synchronizations, constructing complex data structures, and building high-level frameworks. In this paper, we focus on the evaluation and analysis of integer sum reduction, a common data parallel primitive. We convert the sequential reduction into parallel OpenCL implementations on the CPU. We also develop three micro kernels, which allow us to understand the relationships between the kernel performance and the operations involved in reduction. The results of the micro kernels show that increasing the work-group size linearly can linearly improve the kernel performance. There is a sweet spot in the relationship between the work-group size and barrier synchronization overhead. The performance of the atomics over local memory are not sensitive to the work-group size. The sum reduction kernel with vectorized memory accesses can improve the performance of the baseline kernel for a wide range of work-group sizes. However, the vectorization efficiency shrinks with the growing work-group size. We also find that the vendor’s default OpenCL kernel optimization does not improve the kernel performance. On average, disabling the optimization can reduce the execution time of the kernel with vectorized memory accesses by 15%. We attribute the performance drop to the fact that the default kernel optimizations instantiate a large number of atomics over global memory when implicitly vectorizing the kernel computation.

atomics↗

Evaluating LULESH Kernels on OpenCL FPGA

FPGAs are becoming promising heterogeneous computing components for high-performance computing. In this paper, we evaluate the resource utilizations, performance, and performance per watt of our implementations of the LULESH kernels in OpenCL on an Arria10-based FPGA platform. LULESH is a complex proxy application in the CORAL benchmark suite. We choose two representative kernels “CalcFBHourglassForceForElems” and “EvalEOSForElems” from the application in our study. Compared with the baseline implementations, our optimizations improve the performance by a factor of 1.65X and 2.96X for the two kernels on the FPGA, respectively. Using directives for accelerator programming, we also evaluate the performance of the kernels on an Intel Xeon 16-core CPU and an Nvidia K80 GPU. We find that the FPGA, constrained by the memory bandwidth, can perform 1.05X to 3.4X better than the CPU and GPU for small problem sizes. For the first kernel, the performance per watt on the FPGA is 1.59X and 7.1X higher than that on an Intel Xeon 16-core CPU and an Nvidia K80 GPU, respectively. For the second kernel, the performance per watt on the GPU is 1.82X higher than that on the FPGA. However, the performance per watt on the FPGA is 1.77X higher than that on the CPU.

FPGA↗