SEARCH · Engineering Papers
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.
This is the final report of the work on parallel programming patterns that was part of the Center for Programming Models for Scalable Parallel Computing
Fiats provides a platform for research on the training and deployment of neural-network surrogate models for computational science. Fiats also supports exploring, advancing, and combining functional, object-oriented, and parallel programming patterns in Fortran 2023. As such, the Fiats name has dual expansions: “Functional Inference And Training for Surrogates” or “Fortran Inference And Training for Science.” Fiats inference and training procedures are pure and therefore satisfy a language constraint imposed on procedure invocations inside Fortran’s parallel loop construct: do concurrent. Furthermore, the Fiats training procedures are built around a do concurrent parallel reduction. Several compilers can automatically parallelize do concurrent on Central Processing Units (CPUs) or Graphics Processing Units (GPUs). Fiats thus aims to achieve performance portability through standard language mechanisms.
Graphator is a collection of relatively simple sequential programs that generate communication graphs/matrices for commonly occurring patterns in parallel programs. Currently, there is support for five communication patterns: two-dimensional 4-point stencil, four-dimensional 8-point stencil, all-to-alls over sub-communicators, random near-neighbor communication, and near-neighbor communication.
Graphator is a collection of relatively simple sequential programs that generate communication graphs/matrices for commonly occurring patterns in parallel programs. Currently, there is support for five communication patterns: two-dimensional 4-point stencil, four-dimensional 8-point stencil, all-to-alls over sub-communicators, random near-neighbor communication, and near-neighbor communication.
This paper presents a unified approach to unit testing and runtime assertion checking using Fortran 2023. The paper describes the support for our approach in the Julienne framework. Julienne leverages recent Fortran standards to implement object-oriented design patterns, support testing parallel programs, and implement functional programming patterns in order to craft idioms inspired by natural-language expressions. The presented idioms employ novel operators to write expressions that evaluate to a test-diagnosis object encapsulating two components: (1) the test outcome or assertion outcome and (2) an automatically generated diagnostic string. Two other novel aspects of the approach include (1) the ability to enforce assertions inside pure procedures and (2) the ability to output rich diagnostic information inside pure procedures during error termination when assertions fail. The latter capability mitigates against a reason that Fortran programmers commonly cite for not writing pure procedures: difficulty obtaining useful program output inside pure procedures when debugging code. This paper demonstrates how the adoption of the proposed idioms leads naturally to a unifying theme across two otherwise disparate technologies: unit testing and runtime assertion checking. Finally, this paper describes the usage of the Julienne testing framework for writing unit tests and assertions in the Matcha high-performance computing application and the Fiats deep learning library.
Parallel patterns, views, and spaces are promising abstractions to capture the programmer's intent as well as the contextual information that can be used by an underlying runtime to efficiently map software to parallel hardware. These abstractions can be valuable in cases where an algorithm must accommodate requirements of code and performance portability across hardware architectures and vendor programming models. Kokkos is a parallel programming model for host- and accelerator architectures that relies on these abstractions and targets these requirements. It consists of a pure C++ interface, a specification, and a programming library. The programming library exposes patterns and types and maps them to an underlying abstract machine model. The abstract machine model offers a generic view of parallel hardware. While Kokkos is gaining popularity in large-scale HPC applications at some DOE laboratories, we believe that the implemented concepts are of interest to a broader audience including academia as they may contribute to a generic, vendor, and architecture-independent education of parallel programming. In this work, we give an insight into the design considerations of this programming model and list important abstractions. Further, we document best practices obtained from giving virtual classes on Kokkos and give pointers to resources that the reader may consider valuable for a lecture on generic parallel programming for students with preexisting knowledge on this matter.
These are the slides for a presentation on programming models in HPC, at the Los Alamos National Laboratory's Parallel Computing Summer School. The following topics are covered: Flynn's Taxonomy of computer architectures; single instruction single data; single instruction multiple data; multiple instruction multiple data; address space organization; definition of Trinity (Intel Xeon-Phi is a MIMD architecture); single program multiple data; multiple program multiple data; ExMatEx workflow overview; definition of a programming model, programming languages, runtime systems; programming model and environments; MPI (Message Passing Interface); OpenMP; Kokkos (Performance Portable Thread-Parallel Programming Model); Kokkos abstractions, patterns, policies, and spaces; RAJA, a systematic approach to node-level portability and tuning; overview of the Legion Programming Model; mapping tasks and data to hardware resources; interoperability: supporting task-level models; Legion S3D execution and performance details; workflow, integration of external resources into the programming model.
This paper presents ideas for using coordinate-free numerics in modern Fortran to achieve code flexibility in the partial differential equation (PDE) domain. We also show how Fortran, over the last few decades, has changed to become a language well-suited for state-of-the-art software development. Fortran’s new coarray distributed data structure, the language’s class mechanism, and its side-effect-free, pure procedure capability provide the scaffolding on which we implement HPC software. These features empower compilers to organize parallel computations with efficient communication. We present some programming patterns that support asynchronous evaluation of expressions comprised of parallel operations on distributed data. We implemented these patterns using coarrays and the message passing interface (MPI). We compared the codes’ complexity and performance. The MPI code is much more complex and depends on external libraries. The MPI code on Cray hardware using the Cray compiler is 1.5–2 times faster than the coarray code on the same hardware. The Intel compiler implements coarrays atop Intel’s MPI library with the result apparently being 2–2.5 times slower than manually coded MPI despite exhibiting nearly linear scaling efficiency. As compilers mature and further improvements to coarrays comes in Fortran 2015, we expect this performance gap to narrow.
The manycore revolution can be characterized by increasing thread counts, decreasing memory per thread, and diversity of continually evolving manycore architectures. High performance computing (HPC) applications and libraries must exploit increasingly finer levels of parallelism within their codes to sustain scalability on these devices. We found that a major obstacle to performance portability is the diverse and conflicting set of constraints on memory access patterns across devices. Contemporary portable programming models address manycore parallelism (e.g., OpenMP, OpenACC, OpenCL) but fail to address memory access patterns. The Kokkos C++ library enables applications and domain libraries to achieve performance portability on diverse manycore architectures by unifying abstractions for both fine-grain data parallelism and memory access patterns. In this paper we describe Kokkos’ abstractions, summarize its application programmer interface (API), present performance results for unit-test kernels and mini-applications, and outline an incremental strategy for migrating legacy C++ codes to Kokkos. Furthermore, the Kokkos library is under active research and development to incorporate capabilities from new generations of manycore architectures, and to address a growing list of applications and domain libraries.
According to the article, architectures are rapidly evolving, and exascale machines are expected to offer billion-way concurrency. We need to rethink algorithms, languages and programming models among other components in order to migrate large scale applications and explore parallelism on these machines. Although directive-based programming models allow programmers to worry less about programming and more about science, expressing complex parallel patterns in these models can be a daunting task especially when the goal is to match the performance that the hardware platforms can offer. One such pattern is wavefront. This paper extensively studies a wavefront-based miniapplication for Denovo, a production code for nuclear reactor modeling. We parallelize the Koch–Baker–Alcouffe (KBA) parallel-wavefront sweep algorithm in the main kernel of Minisweep (the miniapplication) using CUDA 9.0, OpenMP 4.0 (SIMD) and OpenACC 2.6. Our OpenACC implementation running on NVIDIA’s next-generation Volta GPU boasts an 85.06x speedup over serial code, which is larger than CUDA’s 83.72x speedup over the same serial implementation. We also explore the scalability of our solution using MPI to decompose our simulation domain, allowing us to run on many nodes and accelerators present in state-of-the-art HPC systems. Our parallelization effort across platforms also motivated us to define an abstract parallelism model that is architecture independent, with a goal of creating software abstractions that can be used by applications employing the wavefront sweep motif.
Architectures are rapidly evolving, and exascale machines are expected to offer billion-way concurrency. We need to rethink algorithms, languages and programming models among other components in order to migrate large scale applications and explore parallelism on these machines. Although directive-based programming models allow programmers to worry less about programming and more about science, expressing complex parallel patterns in these models can be a daunting task especially when the goal is to match the performance that the hardware platforms can offer. One such pattern is wavefront. This paper extensively studies a wavefront-based miniapplication for Denovo, a production code for nuclear reactor modeling. We parallelize the Koch-Baker-Alcouffe (KBA) parallel-wavefront sweep algorithm in the main kernel of Minisweep (the miniapplication) using CUDA, OpenMP and OpenACC. Our OpenACC implementation running on NVIDIA's next-generation Volta GPU boasts an 85.06x speedup over serial code, which is larger than CUDA's 83.72x speedup over the same serial implementation. Our experimental platform includes SummitDev, an ORNL representative architecture of the upcoming Summit supercomputer. Our parallelization effort across platforms also motivated us to define an abstract parallelism model that is architecture independent, with a goal of creating software abstractions that can be used by applications employing the wavefront sweep motif.
The accompanying poster to this short paper presents a combination of reverse mode AD and formal methods to enable efficient differentiation of (or backpropagation through) shared-memory parallel code. Compared to the state of the art, our approach can more often avoid the need for atomic updates or private data copies during the parallel derivative computation, even in the presence of unstructured or data-dependent data access patterns. This is achieved by gathering information about the memory access patterns from the input program, which is assumed to be correctly parallelized. This information is then used to build a model of assertions in a theorem prover, which can be used to check the safety of shared memory accesses during the parallel derivative computation
We report that replica-exchange molecular dynamics (REMD) has been proven to efficiently improve the convergence of free energy perturbation (FEP) calculations involving considerable reorganization of their surrounding. We previously introduced the FEP/(λ,H)-REMD algorithm for ligand binding, in which replicas along the alchemical thermodynamic coupling axis λ were expanded as a series of Hamiltonian boosted replicas along a second axis to form a two-dimensional (2D) replica-exchange exchange map [Jiang, W.; Roux, B., J. Chem. Theory Comput. 2010, 6 (9), 2559-2565]. Aiming to achieve a similar performance at a lower computational cost, we propose here a modified version of this algorithm in which only the end-states along the alchemical axis are augmented by boosted replicas. The reduced FEP/(λ,H)-REMD method with one-dimensional (1D) unbiased alchemical thermodynamic coupling axis λ is implemented on the basis of generic multiple copy algorithm (MCA) module of the biomolecular simulation program NAMD. The flexible MCA framework of NAMD enables a user to design customized replica-exchange patterns through Tcl scripting in the context of a highly parallelized simulation program without touching the source code. Two Hamiltonian tempering boosting scheme were examined with the new algorithm: a first one based on potential energy rescaling of a pre-identified “solute”, and a second one via the introduction of flattening torsional free energy barriers. As two illustrative examples with reliable experiment data, the absolute binding free energies of pxylene and n-butylbenzene to the nonpolar cavity of the L99A mutant of T4 lysozyme were calculated. Lastly, the tests demonstrate that the new protocol efficiently enhances the sampling of torsional motions for backbone and side chains around the binding pocket and accelerates the convergence of the free energy computations.
The enhanced central and eastern Pacific SST warming and the associated ocean processes under global warming are investigated using the ocean component of the Community Earth System Model (CESM), Parallel Ocean Program version 2 (POP2). The tropical SST warming pattern in the coupled CESM can be faithfully reproduced by the POP2 forced with surface fluxes computed using the aerodynamic bulk formula. By prescribing the wind stress and/or wind speed through the bulk formula, the effects of wind stress change and/or the wind-evaporation-SST (WES) feedback are isolated and their linearity is evaluated in this ocean-alone setting. Result shows that, although the weakening of the equatorial easterlies contributes positively to the El Niño-like SST warming, 80% of which can be simulated by the POP2 without considering the effects of wind change in both mechanical and thermodynamic fluxes. This result points to the importance of the air-sea thermal interaction and the relative feebleness of the ocean dynamical process in the El Niño-like equatorial Pacific SST response to global warming. On the other hand, the wind stress change is found to play a dominant role in the oceanic response in the tropical Pacific, accounting for most of the changes in the equatorial ocean current system and thermal structures, including the weakening of the surface westward currents, the enhancement of the near-surface stratification and the shoaling of the equatorial thermocline. Interestingly, greenhouse gas warming in the absence of wind stress change and WES feedback also contributes substantially to the changes at the subsurface equatorial Pacific. Further, this warming impact can be largely replicated by an idealized ocean experiment forced by a uniform surface heat flux, whereby, arguably, a purest form of oceanic dynamical thermostat is revealed.
Recent work has shown that program analysis techniques to select meaningful code features of programs are important in the task of deciding the best compiler optimizations. Although, there are many successful state-of-the-art program analysis techniques, they often do not provide a simple method to extract the most expressive information about loops, especially when a target program is computationally intensive with complex loops and data dependencies. In this paper, we introduce a static technique to characterize a program using a pattern-driven system named HERCULES. This characterization technique not only helps a user to understand programs by searching pattern-of-interests, but also can be used for a predictive model that effectively selects the proper compiler optimizations. We formulated 35 loop patterns, then evaluated our characterization technique by comparing the predictive models constructed using HERCULES to three other state-of-the-art characterization methods. We show that our models outperform three state-of-the-art program characterization techniques on two multicore systems in selecting the best optimization combination from a given loop transformation space. We achieved up to 67% of the best possible speedup achievable with the optimization search space we evaluated.
Genomic datasets are growing dramatically as the cost of sequencing continues to decline and small sequencing devices become available. Enormous community databases store and share these data with the research community, but some of these genomic data analysis problems require large-scale computational platforms to meet both the memory and computational requirements. These applications differ from scientific simulations that dominate the workload on high-end parallel systems today and place different requirements on programming support, software libraries and parallel architectural design. For example, they involve irregular communication patterns such as asynchronous updates to shared data structures. We consider several problems in high-performance genomics analysis, including alignment, profiling, clustering and assembly for both single genomes and metagenomes. We identify some of the common computational patterns or ‘motifs’ that help inform parallelization strategies and compare our motifs to some of the established lists, arguing that at least two key patterns, sorting and hashing, are missing. This article is part of a discussion meeting issue ‘Numerical algorithms for high-performance computational science’.
DIY2 is a programming model and runtime for block-parallel analytics on distributed-memory machines. Its main abstraction is block-structured data parallelism: data are decomposed into blocks; blocks are assigned to processing elements (processes or threads); computation is described as iterations over these blocks, and communication between blocks is defined by reusable patterns. By expressing computation in this general form, the DIY2 runtime is free to optimize the movement of blocks between slow and fast memories (disk and flash vs. DRAM) and to concurrently execute blocks residing in memory with multiple threads. This enables the same program to execute in-core, out-of-core, serial, parallel, single-threaded, multithreaded, or combinations thereof. This paper describes the implementation of the main features of the DIY2 programming model and optimizations to improve performance. DIY2 is evaluated on benchmark test cases to establish baseline performance for several common patterns and on larger complete analysis codes running on large-scale HPC machines.
Performance models are powerful instruments for understanding the performance of parallel systems and uncovering their bottlenecks. Already during system design, performance models can help ponder alternative development options. However, creating a performance model – whether theoretically or empirically – for an entire application that does not exist yet is challenging. In this paper, we propose to generate performance models of full programs from performance models of their components using formal composition operators derived from parallel design patterns. As long as the design of the overall system follows such a pattern, its performance model can be predicted with reasonable accuracy without an actual implementation. In conclusion, we demonstrate our approach with design patterns of varying complexity, including pipeline, task pool, and eventually MapReduce, which is representative of a broad class of data-analytics applications.