arXiv subjects

Hartmut Kaiser

Publications and source records attributed to Hartmut Kaiser.

49 records · Page 3Linked to original sources

arXiv2020

Implementing Software Resiliency in HPX for Extreme Scale Computing

Exceptions and errors occurring within mission critical applications due to hardware failures have a high cost. With the emerging Next Generation Platforms (NGPs), the rate of hardware failures will invariably increase. Therefore, designing our applications to be resilient is a critical concern in order to retain the reliability of results while meeting the constraints on power budgets. In this paper, we implement software resilience in HPX, an Asynchronous Many-Task Runtime system. We implement two resiliency APIs that we expose to the application developers, namely task replication and task replay. Task replication repeats a task n-times and executes them asynchronously. Task replay will reschedule a task up to n-times until a valid output is returned. Furthermore, we introduce an API that allows the application to verify the returned result with a user provided predicate. We test the APIs with both artificial workloads and a dataflow based stencil application. We demonstrate that only minor overheads are incurred when utilizing these resiliency features for work loads where the task size is greater than 200 $μ$s. We also show that most of the added execution time arises from the replay or replication of the tasks themselves and not by the implementation of the APIs.

cs.DC↗

arXiv2020

Supporting OpenMP 5.0 Tasks in hpxMP -- A study of an OpenMP implementation within Task Based Runtime Systems

OpenMP has been the de facto standard for single node parallelism for more than a decade. Recently, asynchronous many-task runtime (AMT) systems have increased in popularity as a new programming paradigm for high performance computing applications. One of the major challenges of this new paradigm is the incompatibility of the OpenMP thread model and other AMTs. Highly optimized OpenMP-based libraries do not perform well when coupled with AMTs because the threading of both libraries will compete for resources. This paper is a follow-up paper on the fundamental implementation of hpxMP, an implementation of the OpenMP standard which utilizes the C++ standard library for Parallelism and Concurrency (HPX) to schedule and manage tasks. In this paper, we present the implementation of task features, e.g. taskgroup, task depend, and task_reduction, of the OpenMP 5.0 standard and optimization of the #pragma omp parallel for pragma. We use the daxpy benchmark, the Barcelona OpenMP Tasks Suite, Parallel research kernels, and OpenBLAS benchmarks to compare the different OpenMp implementations: hpxMP, llvm-OpenMP, and GOMP.

cs.DC↗

arXiv2019

From Piz Daint to the Stars: Simulation of Stellar Mergers using High-Level Abstractions

We study the simulation of stellar mergers, which requires complex simulations with high computational demands. We have developed Octo-Tiger, a finite volume grid-based hydrodynamics simulation code with Adaptive Mesh Refinement which is unique in conserving both linear and angular momentum to machine precision. To face the challenge of increasingly complex, diverse, and heterogeneous HPC systems, Octo-Tiger relies on high-level programming abstractions. We use HPX with its futurization capabilities to ensure scalability both between nodes and within, and present first results replacing MPI with libfabric achieving up to a 2.8x speedup. We extend Octo-Tiger to heterogeneous GPU-accelerated supercomputers, demonstrating node-level performance and portability. We show scalability up to full system runs on Piz Daint. For the scenario's maximum resolution, the compute-critical parts (hydrodynamics and gravity) achieve 68.1% parallel efficiency at 2048 nodes.

cs.DC↗

arXiv2019

An Introduction to hpxMP: A Modern OpenMP Implementation Leveraging HPX, An Asynchronous Many-Task System

Asynchronous Many-task (AMT) runtime systems have gained increasing acceptance in the HPC community due to the performance improvements offered by fine-grained tasking runtime systems. At the same time, C++ standardization efforts are focused on creating higher-level interfaces able to replace OpenMP or OpenACC in modern C++ codes. These higher level functions have been adopted in standards conforming runtime systems such as HPX, giving users the ability to simply utilize fork-join parallelism in their own codes. Despite innovations in runtime systems and standardization efforts users face enormous challenges porting legacy applications. Not only must users port their own codes, but often users rely on highly optimized libraries such as BLAS and LAPACK which use OpenMP for parallization. Current efforts to create smooth migration paths have struggled with these challenges, especially as the threading systems of AMT libraries often compete with the treading system of OpenMP. To overcome these issues, our team has developed hpxMP, an implementation of the OpenMP standard, which utilizes the underlying AMT system to schedule and manage tasks. This approach leverages the C++ interfaces exposed by HPX and allows users to execute their applications on an AMT system without changing their code. In this work, we compare hpxMP with Clang's OpenMP library with four linear algebra benchmarks of the Blaze C++ library. While hpxMP is often not able to reach the same performance, we demonstrate viability for providing a smooth migration for applications but have to be extended to benefit from a more general task based programming model.

cs.DC↗

arXiv2018

Integration of CUDA Processing within the C++ library for parallelism and concurrency (HPX)

Experience shows that on today's high performance systems the utilization of different acceleration cards in conjunction with a high utilization of all other parts of the system is difficult. Future architectures, like exascale clusters, are expected to aggravate this issue as the number of cores are expected to increase and memory hierarchies are expected to become deeper. One big aspect for distributed applications is to guarantee high utilization of all available resources, including local or remote acceleration cards on a cluster while fully using all the available CPU resources and the integration of the GPU work into the overall programming model. For the integration of CUDA code we extended HPX, a general purpose C++ run time system for parallel and distributed applications of any scale, and enabled asynchronous data transfers from and to the GPU device and the asynchronous invocation of CUDA kernels on this data. Both operations are well integrated into the general programming model of HPX which allows to seamlessly overlap any GPU operation with work on the main cores. Any user defined CUDA kernel can be launched on any (local or remote) GPU device available to the distributed application. We present asynchronous implementations for the data transfers and kernel launches for CUDA code as part of a HPX asynchronous execution graph. Using this approach we can combine all remotely and locally available acceleration cards on a cluster to utilize its full performance capabilities. Overhead measurements show, that the integration of the asynchronous operations (data transfer + launches of the kernels) as part of the HPX execution graph imposes no additional computational overhead and significantly eases orchestrating coordinated and concurrent work on the main cores and the used GPU devices.

cs.DC↗

arXiv2018

Asynchronous Execution of Python Code on Task Based Runtime Systems

Despite advancements in the areas of parallel and distributed computing, the complexity of programming on High Performance Computing (HPC) resources has deterred many domain experts, especially in the areas of machine learning and artificial intelligence (AI), from utilizing performance benefits of such systems. Researchers and scientists favor high-productivity languages to avoid the inconvenience of programming in low-level languages and costs of acquiring the necessary skills required for programming at this level. In recent years, Python, with the support of linear algebra libraries like NumPy, has gained popularity despite facing limitations which prevent this code from distributed runs. Here we present a solution which maintains both high level programming abstractions as well as parallel and distributed efficiency. Phylanx, is an asynchronous array processing toolkit which transforms Python and NumPy operations into code which can be executed in parallel on HPC resources by mapping Python and NumPy functions and variables into a dependency tree executed by HPX, a general purpose, parallel, task-based runtime system written in C++. Phylanx additionally provides introspection and visualization capabilities for debugging and performance analysis. We have tested the foundations of our approach by comparing our implementation of widely used machine learning algorithms to accepted NumPy standards.

cs.PL↗

arXiv2017

HPX Smart Executors

The performance of many parallel applications depends on loop-level parallelism. However, manually parallelizing all loops may result in degrading parallel performance, as some of them cannot scale desirably to a large number of threads. In addition, the overheads of manually tuning loop parameters might prevent an application from reaching its maximum parallel performance. We illustrate how machine learning techniques can be applied to address these challenges. In this research, we develop a framework that is able to automatically capture the static and dynamic information of a loop. Moreover, we advocate a novel method by introducing HPX smart executors for determining the execution policy, chunk size, and prefetching distance of an HPX loop to achieve higher possible performance by feeding static information captured during compilation and runtime-based dynamic information to our learning model. Our evaluated execution results show that using these smart executors can speed up the HPX execution process by around 12%-35% for the Matrix Multiplication, Stream and $2D$ Stencil benchmarks compared to setting their HPX loop's execution policy/parameters manually or using HPX auto-parallelization techniques.

cs.DC↗

arXiv2017

Redesigning OP2 Compiler to Use HPX Runtime Asynchronous Techniques

Maximizing parallelism level in applications can be achieved by minimizing overheads due to load imbalances and waiting time due to memory latencies. Compiler optimization is one of the most effective solutions to tackle this problem. The compiler is able to detect the data dependencies in an application and is able to analyze the specific sections of code for parallelization potential. However, all of these techniques provided with a compiler are usually applied at compile time, so they rely on static analysis, which is insufficient for achieving maximum parallelism and producing desired application scalability. One solution to address this challenge is the use of runtime methods. This strategy can be implemented by delaying certain amount of code analysis to be done at runtime. In this research, we improve the parallel application performance generated by the OP2 compiler by leveraging HPX, a C++ runtime system, to provide runtime optimizations. These optimizations include asynchronous tasking, loop interleaving, dynamic chunk sizing, and data prefetching. The results of the research were evaluated using an Airfoil application which showed a 40-50% improvement in parallel performance.

cs.DC↗

arXiv2017

A Load-Balanced Parallel and Distributed Sorting Algorithm Implemented with PGX.D

Sorting has been one of the most challenging studied problems in different scientific researches. Although many techniques and algorithms have been proposed on the theory of having efficient parallel sorting implementation, however achieving desired performance on different types of the architectures with large number of processors is still a challenging issue. Maximizing parallelism level in applications can be achieved by minimizing overheads due to load imbalance and waiting time due to memory latencies. In this paper, we present a distributed sorting algorithm implemented in PGX.D, a fast distributed graph processing system, which outperforms the Spark's distributed sorting implementation by around 2x-3x by hiding communication latencies and minimizing unnecessary overheads. Furthermore, it shows that the proposed PGX.D sorting method handles dataset containing many duplicated data entries efficiently and always results in keeping balanced workloads for different input data distribution types.

cs.DC↗

arXiv2012

Neutron Star Evolutions using Tabulated Equations of State with a New Execution Model

The addition of nuclear and neutrino physics to general relativistic fluid codes allows for a more realistic description of hot nuclear matter in neutron star and black hole systems. This additional microphysics requires that each processor have access to large tables of data, such as equations of state, and in large simulations the memory required to store these tables locally can become excessive unless an alternative execution model is used. In this work we present relativistic fluid evolutions of a neutron star obtained using a message driven multi-threaded execution model known as ParalleX. These neutron star simulations would require substantial memory overhead dedicated entirely to the equation of state table if using a more traditional execution model. We introduce a ParalleX component based on Futures for accessing large tables of data, including out-of-core sized tables, which does not require substantial memory overhead and effectively hides any increased network latency.

cs.DC↗

arXiv2011

Adaptive Mesh Refinement for Astrophysics Applications with ParalleX

Several applications in astrophysics require adequately resolving many physical and temporal scales which vary over several orders of magnitude. Adaptive mesh refinement techniques address this problem effectively but often result in constrained strong scaling performance. The ParalleX execution model is an experimental execution model that aims to expose new forms of program parallelism and eliminate any global barriers present in a scaling-impaired application such as adaptive mesh refinement. We present two astrophysics applications using the ParalleX execution model: a tabulated equation of state component for neutron star evolutions and a cosmology model evolution. Performance and strong scaling results from both simulations are presented. The tabulated equation of state data are distributed with transparent access over the nodes of the cluster. This allows seamless overlapping of computation with the latencies introduced by the remote access to the table. Because of the expected size increases to the equation of state table, this type of table partitioning for neutron star simulations is essential while the implementation is greatly simplified by ParalleX semantics.

cs.DC↗

arXiv2011

An Application Driven Analysis of the ParalleX Execution Model

Exascale systems, expected to emerge by the end of the next decade, will require the exploitation of billion-way parallelism at multiple hierarchical levels in order to achieve the desired sustained performance. The task of assessing future machine performance is approached by identifying the factors which currently challenge the scalability of parallel applications. It is suggested that the root cause of these challenges is the incoherent coupling between the current enabling technologies, such as Non-Uniform Memory Access of present multicore nodes equipped with optional hardware accelerators and the decades older execution model, i.e., the Communicating Sequential Processes (CSP) model best exemplified by the message passing interface (MPI) application programming interface. A new execution model, ParalleX, is introduced as an alternative to the CSP model. In this paper, an overview of the ParalleX execution model is presented along with details about a ParalleX-compliant runtime system implementation called High Performance ParalleX (HPX). Scaling and performance results for an adaptive mesh refinement numerical relativity application developed using HPX are discussed. The performance results of this HPX-based application are compared with a counterpart MPI-based mesh refinement code. The overheads associated with HPX are explored and hardware solutions are introduced for accelerating the runtime system.

cs.DC↗

arXiv2011

Improving the scalability of parallel N-body applications with an event driven constraint based execution model

The scalability and efficiency of graph applications are significantly constrained by conventional systems and their supporting programming models. Technology trends like multicore, manycore, and heterogeneous system architectures are introducing further challenges and possibilities for emerging application domains such as graph applications. This paper explores the space of effective parallel execution of ephemeral graphs that are dynamically generated using the Barnes-Hut algorithm to exemplify dynamic workloads. The workloads are expressed using the semantics of an Exascale computing execution model called ParalleX. For comparison, results using conventional execution model semantics are also presented. We find improved load balancing during runtime and automatic parallelism discovery improving efficiency using the advanced semantics for Exascale computing.

cs.DC↗