Particle-based simulations and point-cloud applications generate massive, irregular datasets that challenge storage, I/O, and real-time analytics. Traditional compression techniques struggle with irregular particle distributions and GPU architectural constraints, often resulting in limited throughput and suboptimal compression ratios. In this paper, we present GPZ, a high-performance, error-bounded lossy compressor designed specifically for large-scale particle data on modern GPUs. GPZ employs a novel four-stage parallel pipeline that synergistically balances high compression efficiency with the architectural demands of massively parallel hardware. We introduce a suite of targeted optimizations for computation, memory access, and GPU occupancy that enable GPZ to achieve near-hardware-limit throughput. We conduct an extensive evaluation on three distinct GPU architectures (workstation, data center, and edge) using six large-scale, real-world scientific datasets from four distinct domains. The results demonstrate that GPZ consistently and significantly outperforms four state-of-the-art GPU compressors, delivering up to 8x higher end-to-end throughput while achieving superior compression ratios and data quality.

This paper formalizes an emerging learning paradigm that uses a trained model as a reference to guide and enhance the training of a target model through strategic data selection or weighting, named **model steering**. While ad-hoc methods have been used in various contexts, including the training of large foundation models, its underlying principles remain insufficiently understood, leading to sub-optimal performance. In this work, we propose a theory-driven framework for model steering called **DRRho risk minimization**, which is rooted in Distributionally Robust Optimization (DRO). Through a generalization analysis, we provide theoretical insights into why this approach improves generalization and data efficiency compared to training without a reference model. To the best of our knowledge, this is the first time such theoretical insights are provided for the new learning paradigm, which significantly enhance our understanding and practice of model steering. Building on these insights and the connection between contrastive learning and DRO, we introduce a novel method for Contrastive Language-Image Pretraining (CLIP) with a reference model, termed DRRho-CLIP. Extensive experiments validate the theoretical insights, reveal a superior scaling law compared to CLIP without a reference model, and demonstrate its strength over existing heuristic approaches. Code is released at [github.com/Optimization-AI/DRRho-CLIP](https://github.com/Optimization-AI/DRRho-CLIP)

Error-bounded lossy compression is one of the most efficient solutions to reduce the volume of scientific data. For lossy compression, progressive decompression and random-access decompression are critical features that enable on-demand data access and flexible analysis workflows. However, these features can severely degrade compression quality and speed. To address these limitations, we propose a novel streaming compression framework that supports both progressive decompression and random-access decompression while maintaining high compression quality and speed. Our contributions are three-fold: (1) we design the first compression framework that simultaneously enables both progressive decompression and random-access decompression; (2) we introduce a hierarchical partitioning strategy to enable both streaming features, along with a hierarchical prediction mechanism that mitigates the impact of partitioning and achieves high compression quality—even comparable to state-of-the-art (SOTA) non-streaming compressor SZ3; and (3) our framework delivers high compression and decompression speed, up to 6.7 × faster than SZ3.

E3SM [3], which rely on highly specialized couplers to facilitate the exchange of data between participant models.Although domain-specific couplers, like E3SM's CPL7, enable highly efficient data transfer, they are largely inflexible and cannot be utilized with other models outside of their ecosystem.To help domain scientists integrate various users' models, the NSF's Cyberwater framework [2] is built to use a data-exchange service to connect distinct models that execute in distributed computing systems.As a part of the framework, the Data Exchange Service (DES) is a web service-based coupler that allows the exchange of data between scientific models through a generic service component.The Data Exchange Service promotes interoperability among models, with the exception that the models must adhere to sharing certain units of measurement.Developers are able to build such a coupling service, however, it is a challenging task to verify the service's reliability and functionality.Considering the criticality of the correctness of scientific computing simulations, rigorous testing methods are essential to ensure data is handled correctly and to certify the reliability of the DES Controller, as coupling logic resides in this component.This paper is organized as follows.Section 2 provides related work.Section 3 describes our testing methodology.Section 4 and Section 5 show how we apply rigorous specification and testing to the DES Controller, with results in Section 5. Conclusion and future work are given in Section 6.

Learned Index, which utilizes effective machine learning models to accelerate locating sorted data positions, has gained increasing attention in many big data scenarios. Using efficient learned models, the learned indexes build large nodes and flat structures, thereby greatly improving the performance. However, most of the state-of-the-art learned indexes are designed for DRAM, and there is hence an urgent need to enable high-performance learned indexes for emerging Non-Volatile Memory (NVM). In this article, we first evaluate and analyze the performance of the existing learned indexes on NVM. We discover that these learned indexes encounter severe write amplification and write performance degradation due to the requirements of maintaining large sorted/semi-sorted data nodes. To tackle the problems, we propose a novel three-tiered architecture of write-optimized persistent learned index, which is named WIPE , by adopting unsorted fine-granularity data nodes to achieve high write performance on NVM. Thereinto, we devise a new root node construction algorithm to accelerate searching numerous small data nodes. The algorithm ensures stable flat structure and high read performance in large-size datasets by introducing an intermediate layer (i.e., index nodes) and achieving accurate prediction of index node positions from the root node. Our extensive experiments on Intel DCPMM show that WIPE can improve write throughput and read throughput by up to 3.9× and 7×, respectively, compared to the state-of-the-art learned indexes. Also, WIPE can recover from a system crash in ∼ 18 ms. WIPE is free as an open-source software package. 1

The CyberWater project is to build an open-data open-model framework for easy and incremental integration of heterogeneous data sources and diverse scientific models across disciplines in the broad water domain. The CyberWater framework extends the open-data open-model framework called Meta-Scientific-Modeling (MSM) that provides a system-wide data and model integration platform. On top of MSM, the CyberWater framework provides a set of toolkits, and external system integration engines, to further facilitate users' scientific modeling and collaboration across disciplines. For example, the developed generic model agent toolkit enables users to integrate their computational models into CyberWater via graphical user interface configuration without coding, which further simplifies the data and model integration and model coupling. CyberWater adopts a graphical scientific workflow system, VisTrails, ensuring data provenance and reproducible computing. CyberWater supports novel access to high-performance computing resources on demand for users' computational expensive model tasks. We demonstrate merits of CyberWater by a use case of hydrologic modeling workflow.

Workflow management systems (WMSs) are commonly used to organize/automate sequences of tasks as workflows to accelerate scientific discoveries. During complex workflow modeling, a local interactive workflow environment is desirable, as users usually rely on their rich, local environments for fast prototyping and refinements before they consider using more powerful computing resources. However, existing WMSs do not simultaneously support local interactive workflow environments and HPC resources. In this paper, we present an on-demand access mechanism to remote HPC resources from desktop/laptop-based workflow management software to compose, monitor and analyze scientific workflows in the CyberWater project. Cyber-Water is an open-data and open-modeling software framework for environmental and water communities. In this work, we extend the open-model, open-data design of CyberWater with on-demand HPC accessing capacity. In particular, we design and implement the LaunchAgent library, which can be integrated into the local desktop environment to allow on-demand usage of remote resources for hydrology-related workflows. LaunchAgent manages authentication to remote resources, prepares the computationally-intensive or data-intensive tasks as batch jobs, submits jobs to remote resources, and monitors the quality of services for the users. LaunchAgent interacts seamlessly with other existing components in CyberWater, which is now able to provide advantages of both feature-rich desktop software experience and increased computation power through on-demand HPC/Cloud usage. In our evaluations, we demonstrate how a hydrology workflow that consists of both local and remote tasks can be constructed and show that the added on-demand HPC/Cloud usage helps speeding up hydrology workflows while allowing intuitive workflow configurations and execution using a desktop graphical user interface.

Performance modeling of parallel applications on distributed memory systems is a challenging task due to the effects of CPU speed, memory access time, and communication cost. In this paper, we propose a simple and intuitive graphical model, which extends the widely used Roofline performance model to include the communication cost in addition to the memory access time and the peak CPU performance. This new performance model inherits the simplicity of the original Roofline model and enables performance evaluation on a third dimension of communication performance. Such a model will greatly facilitate and expedite the analysis, development and optimization of parallel programs on high-end computer systems. We empirically validate the extended new Roofline model usingfl oating-point-computation-bound, memory-bound, and communication-bound applications. Three distinct high-end computing platforms have been tested: 1) high performance computing (HPC) systems, 2) high throughput computing systems, and 3) cloud computing systems. Our experimental results with four different parallel applications show that the new model can approximately evaluate the performance of different programs on various distributed-memory systems. Furthermore, the extended new model is able to provide insight into how the problem size can affect the upper bound performance of parallel applications, which is a special property revealed by the new dimension of communication cost analysis.

Summary Nowadays, we have entered the era of big data. In the area of high performance computing, large‐scale simulations can generate huge amounts of data with potentially critical information. However, these data are usually saved in intermediate files and are not instantly visible until advanced data analytics techniques are applied after reading all simulation data from persistent storages (eg, local disks or a parallel file system). This approach puts users in a situation where they spend long time on waiting for running simulations while not knowing the status of the running job. In this paper, we build a new computational framework to couple scientific simulations with multi‐step machine learning processes and in‐situ data visualizations. We also design a new scalable simulation‐time clustering algorithm to automatically detect fluid flow anomalies. This computational framework is built upon different software components and provides plug‐in data analysis and visualization functions over complex scientific workflows. With this advanced framework, users can monitor and get real‐time notifications of special patterns or anomalies from ongoing extreme‐scale turbulent flow simulations.

This paper targets an important class of applications that requires combining HPC simulations with data analysis for online or real-time scientific discovery. We use the state-of-the-art parallel-IO and data-staging libraries to build simulation-time data analysis workflows, and conduct performance analysis with real-world applications of computational fluid dynamics (CFD) simulations and molecular dynamics (MD) simulations. Driven by in-depth performance inefficiency analysis, we design an end-to-end application-level approach to eliminating the interlocks and synchronizations existent in the present methods. Our new approach employs both task parallelism and pipeline parallelism to reduce synchronizations effectively. In addition, we design a fully asynchronous, fine-grain, and pipelining runtime system, which is named Zipper. Zipper is a multi-threaded distributed runtime system and executes in a layer below the simulation and analysis applications. To further reduce the simulation application's stall time and enhance the data transfer performance, we design a concurrent data transfer optimization that uses both HPC network and parallel file system for improved bandwidth. The scalability of the Zipper system has been verified by a performance model and various empirical large scale experiments. The experimental results on an Intel multicore cluster as well as a Knight Landing HPC system demonstrate that the Zipper based approach can outperform the fastest state-of-the-art I/O transport library by up to 220% using 13,056 processor cores.

Lattice Boltzmann method (LBM) is an important computational fluid dynamics (CFD) approach to solving the Naiver-Stokes equations and simulating complex fluid flows. LBM is also well known as a memory bound problem and its performance is limited by the memory access time on modern computer systems. In this paper, we design and develop both sequential and parallel memory-aware algorithms to optimize the performance of LBM. The new memory-aware algorithms can enhance data reuses across multiple time steps to further improve the performance of the original and fused LBM. We theoretically analyze the algorithms to provide an insight into how data reuses occur in each algorithm. Finally, we conduct experiments and detailed performance analysis on two different manycore systems. Based on the experimental results, the parallel memory-aware LBM algorithm can outperform the fused LBM by up to 292% on the Intel Haswell system when using 28 cores, and by 302 % on the Intel Skylake system when using 48 cores.

Markov chain usage-based statistical testing has proved sound and effective in providing audit trails of evidence in certifying software-intensive systems.The system end-toend reliability is derived analytically in closed form, following an arc-based Bayesian model.System reliability is represented by an important statistic called single use reliability, and defined as the probability of a randomly selected use being successful.This paper continues our earlier work on a simpler and faster derivation of the single use reliability mean, and proposes a new derivation of the single use reliability variance by applying a well-known theorem and eliminating the need to compute the second moments of arc failure probabilities.Our new results complete a new analysis that could be shown to be simpler, faster, and more direct while also rendering a more intuitive explanation.Our new theory is illustrated with three simple Markov chain usage models with manual derivations and experimental results.

To reduce energy consumption and carbon emission, many data centers have deployed (or anticipate to build) their own renewable-energy power plants. However, the renewable energy (such as wind, tide, and solar energy) has the serious issues of intermittency and variability that prevent the green energy from being utilized effectively in practice. To cope with the issues, new power-supply management policies and workload scheduling algorithms have been designed. However, most existing work focuses on power optimization on computation only. In this paper, we introduce a novel scheme called OptiMatch to optimize the match between the power supply and the user-workload demand for massive storage systems that are mostly powered by renewable energy sources. OptiMatch has a hierarchical architecture, which consists of a number of heterogeneous storage devices. OptiMatch systematically utilizes the performance disparities between heterogeneous storage devices (i.e., performance per watt, IOPS/watt) to split the process for every write request into two stages: an on-line stage and a deferred off-line stage. The deferred off-line requests are used to match the green energy supplies. To maximize green energy utilization and minimize power budget without sacrificing quality of service, the fundamental methodology is to make the aggregate power supplies be proportional to the I/O workload demand at any time. To this end, our OptiMatch employs novel co-design optimizations. (1) We propose a dual-drive power control approach that makes the number of active nodes proportional to the workload demand when the green power supply is insufficient, meanwhile be proportional to the green power supply when green power is sufficient. (2) During periods of insufficient green supplies, we exploit virtualization consolidation schemes which enable a fine-grained power control to minimize the grid budgets. (3) During the periods of sufficient green supplies, we design an intelligent workload scheduling scheme which enables a near-optimal off-line requests assignment to maximize the green utilization. The experimental results demonstrate that the new OptiMatch framework can achieve high green utilization (up to 94.9%) with a minor performance degradation (less than 9.8%).

While many algorithm-based fault tolerance (ABFT) schemes have been proposed to detect soft errors offline in the fast Fourier transform (FFT) after computation finishes, none of the existing ABFT schemes detect soft errors online before the computation finishes. This paper presents an online ABFT scheme for FFT so that soft errors can be detected online and the corrupted computation can be terminated in a much more timely manner. We also extend our scheme to tolerate both arithmetic errors and memory errors, develop strategies to reduce its fault tolerance overhead and improve its numerical stability and fault coverage, and finally incorporate it into the widely used FFTW library - one of the today's fastest FFT software implementations. Experimental results demonstrate that: (1) the proposed online ABFT scheme introduces much lower overhead than the existing offline ABFT schemes; (2) it detects errors in a much more timely manner; and (3) it also has higher numerical stability and better fault coverage.

Markov chain usage-based statistical testing has been around for more than two decades, and proved sound and effective in providing audit trails of evidence in certifying software-intensive systems.The system end-to-end reliability is derived analytically in closed form, following an arc-based Bayesian model.System reliability is represented by an important statistic called single use reliability, and defined as the probability of a randomly selected use being successful.This paper reviews the analytical derivation of the single use reliability mean, and proposes a simpler, faster, and more direct way to compute the expected value that renders an intuitive explanation.The new derivation is illustrated with two examples.

The scope of this paper is to design and implement a scalable QR factorization solver that can deliver the fastest performance for tall and skinny matrices and square matrices on modern supercomputers. The new solver, named scalable universal communication-avoiding QR factorization (suCAQR), introduces a simplified and tuning-less way to realize the communication-avoiding QR factorization algorithm to support matrices of any shapes. The software design includes a mixed usage of physical and logical data layouts, a simplified method of dynamic-root binary-tree reduction, and a dynamic dataflow implementation. Compared with the existing communication avoiding QR factorization implementations, suCAQR has the benefits of being simpler, more general, and more efficient. By balancing the degree of parallelism and the proportion of faster computational kernels, it is able to achieve scalable performance on clusters of multicore nodes. The software essentially combines the strengths of both synchronization-reducing approach and communication-avoiding approach to achieve high performance. Based on the experimental results using 1,024 CPU cores, suCAQR is faster than DPLASMA by up to 30%, and faster than ScaLAPACK by up to 30 times.

Sequence-based software specification is a rigorous method for deriving a formal system model based on informal requirements, through a systematic process called sequence enumeration. Under this process, stimulus (input) sequences are considered in a breadth-first manner, with the expected system response to each sequence given. Not every sequence needs to be further extended by the enumeration rules. The completed specification encodes a Mealy machine and forms a basis for other activities including code development and testing. This paper presents a forward reduction algorithm for sequence-based specification. The need for such an algorithm has been identified by field applications. We used the state machine as an intermediate tool to comprehend and analyze all change impacts resulted from a forward reduction, and used an axiom system for its development. We present the algorithm both mathematically in functional form and procedurally in pseudocode, illustrate it with a symbolic example, and report a larger case study from the published literature in which the algorithm is applied. The algorithm will prove useful and effective in deriving a system-level specification as well as in merging and combining partial work products towards a formal system model in field applications.

Deformable structures are abundant in various domains such as biology, medicine, life sciences, and ocean engineering. Our previous work created a numerical method, named LBM-IB method [1], to solve the fluid-structure interaction (FSI) problems. Our LBM-IB method is particularly suitable for simulating flexible (or elastic) structures immersed in a moving viscous fluid. Fluid-structure interaction problems are well known for their heavy demands on computing resources. Today, it is still challenging to resolve many real-world FSI problems. In order to solve large-scale fluid-structure interactions more efficiently, in this paper, we design a parallel LBM-IB library on shared memory many core architectures. We start from a sequential version, which is extended to two different parallel versions. The paper first introduces the mathematical background of the LBM-IB method, then uses the sequential version as a ground to present our implemented computational kernels and the algorithm. Next, it describes the two parallel programs: an Open MP implementation and a cube-based parallel implementation using Pthreads. The cube-based implementation builds upon our new cube-centric algorithm where all the data are stored in cubes and computations are performed on individual cubes in a data-centric manner. By exploiting better data locality and fine-grain block parallelism, the cube-based parallel implementation is able to outperform the Open MP implementation by up to 53% on 64-core computer systems.

Summary Aiming to fully exploit the computing power of all CPUs and all graphics processing units (GPUs) on hybrid CPU‐GPU systems to solve dense linear algebra problems, we design a class of heterogeneous tile algorithms to maximize the degree of parallelism, to minimize the communication volume, and to accommodate the heterogeneity between CPUs and GPUs. The new heterogeneous tile algorithms are executed upon our decentralized dynamic scheduling runtime system, which schedules a task graph dynamically and transfers data between compute nodes automatically. The runtime system uses a new distributed task assignment protocol to solve data dependencies between tasks without any coordination between processing units. By overlapping computation and communication through dynamic scheduling, we are able to attain scalable performance for the double‐precision Cholesky factorization and QR factorization. Our approach demonstrates a performance comparable to Intel MKL on shared‐memory multicore systems and better performance than both vendor (e.g., Intel MKL) and open source libraries (e.g., StarPU) in the following three environments: heterogeneous clusters with GPUs, conventional clusters without GPUs, and shared‐memory systems with multiple GPUs. Copyright © 2014 John Wiley & Sons, Ltd.

While the growing number of cores per chip allows researchers to solve larger scientific and engineering problems, the parallel efficiency of the deployed parallel software starts to decrease. This unscalability problem happens to both vendor-provided and open-source software and wastes CPU cycles and energy. By expecting CPUs with hundreds of cores to be imminent, we have designed a new framework to perform matrix computations for massively many cores. Our performance analysis on manycore systems shows that the unscalability bottleneck is related to Non-Uniform Memory Access (NUMA): memory bus contention and remote memory access latency. To overcome the bottleneck, we have designed NUMA-aware tile algorithms with the help of a dynamic scheduling runtime system to minimize NUMA memory accesses. The main idea is to identify the data that is, either read a number of times or written once by a thread resident on a remote NUMA node, then utilize the runtime system to conduct data caching and movement between different NUMA nodes. Based on the experiments with QR factorizations, we demonstrate that our framework is able to achieve great scalability on a 48-core AMD Opteron system (e.g., parallel efficiency drops only 3% from one core to 48 cores). We also deploy our framework to an extreme-scale shared-memory SGI machine which has 1024 CPU cores and runs a single Linux operating system image. Our framework continues to scale well, and can outperform the vendor-optimized Intel MKL library by up to 750%.

We present a new approach to utilizing all CPU cores and all GPUs on heterogeneous multicore and multi-GPU systems to support dense matrix computations efficiently. The main idea is that we treat a heterogeneous system as a distributed-memory machine, and use a heterogeneous multi-level block cyclic distribution method to allocate data to the host and multiple GPUs to minimize communication. We design heterogeneous algorithms with hybrid tiles to accommodate the processor heterogeneity, and introduce an auto-tuning method to determine the hybrid tile sizes to attain both high performance and load balancing. We have also implemented a new runtime system and applied it to the Cholesky and QR factorizations. Our approach is designed for achieving four objectives: a high degree of parallelism, minimized synchronization, minimized communication, and load balancing. Our experiments on a compute node (with two Intel Westmere hexa-core CPUs and three Nvidia Fermi GPUs), as well as on up to 100 compute nodes on the Keeneland system, demonstrate great scalability, good load balancing, and efficiency of our approach.

GPU-based heterogeneous clusters continue to draw attention from vendors and HPC users due to their high energy efficiency and much improved single-node computational performance, however, there is little parallel software available that can utilize all CPU cores and all GPUs on the heterogeneous system efficiently. On a heterogeneous cluster, the performance of a GPU (or a compute node) increases in a much faster rate than the performance of the PCI-Express connection (or the interconnection network) such that communication eventually becomes the bottleneck of the entire system. To overcome the bottleneck, we developed a multi-level partitioning and distribution method that guarantees a near-optimal communication volume. We have also extended heterogeneous tile algorithms to work on distributed memory GPU clusters. Our main idea is to execute a serial program and generate hybrid-size tasks, and follow a dataflow programming model to fire the tasks on different compute nodes. We then devised a distributed dynamic scheduling runtime system to schedule tasks, and transfer data between hybrid CPU-GPU compute nodes transparently. The runtime system employs a novel distributed task-assignment protocol to solve data dependencies between tasks without coordination between processing units. The runtime system on each node consists of a number of CPU compute threads, a number of GPU compute threads, a task generation thread, an MPI communication thread, and a CUDA communication thread. By overlapping computation and communication through dynamic scheduling, we are able to attain a high performance of 75 TFlops for Cholesky factorization on the heterogeneous Keeneland system using 100 nodes, each with twelve CPU cores and three GPUs. Moreover, our framework is able to attain high performance on distributed-memory clusters without GPUs, and shared-system multiGPUs.

As tile linear algebra algorithms continue achieving high performance on shared-memory multicore architectures, it is a challenging task to make them scalable on distributed-memory multicore cluster machines. The main contribution of this paper is the extension to the distributed-memory environment of the previous work done by Hadri et al. on Communication- Avoiding QR (CA-QR) factorizations for tall and skinny matrices (initially done on shared-memory multicore systems). The fine granularity of tile algorithms associated with communicationavoiding techniques for the QR factorization presents a high degree of parallelism where multiple tasks can be concurrently executed, computation and communication largely overlapped, and computation steps fully pipelined. A decentralized dynamic scheduler has then been integrated as a runtime system to efficiently schedule tasks across the distributed resources. Our experimental results performed on two clusters (with dual-core and 8-core nodes, respectively) and a Cray XT5 system with 12-core nodes show that the tile CA-QR factorization is able to outperform the de facto ScaLAPACK library by up to 4 times for tall and skinny matrices, and has good scalability on up to 3,072 cores.

This paper proposes an analytical model to estimate the cost of running an affinity-based thread schedule on multicore systems. The model consists of three submodels to evaluate the cost of executing a thread schedule: an affinity-graph submodel, a memory hierarchy submodel, and a cost submodel that characterize programs, machines, and costs respectively. We applied the analytical model to both synthetic and real-world applications. The estimated cost accurately predicts which schedule will provide better performance. Due to the NP-hardness of the scheduling problem, we designed an approximation algorithm to compute near-optimal solutions. We have extended the algorithm to support threads with data dependences. We conducted experiments with a computational fluid dynamics (CFD) kernel and Cholesky factorization on both UMA SMP and NUMA DSM machines. The results show that using the optimized thread schedule can improve the program performance by 25% to 400%, demonstrating that our method for determining an optimized thread schedule for multicore systems is efficient and practical.

This paper presents a dynamic task scheduling approach to executing dense linear algebra algorithms on multicore systems (either shared-memory or distributed-memory). We use a task-based library to replace the existing linear algebra subroutines such as PBLAS to transparently provide the same interface and computational function as the ScaLAPACK library. Linear algebra programs are written with the task-based library and executed by a dynamic runtime system. We mainly focus our runtime system design on the metric of performance scalability. We propose a distributed algorithm to solve data dependences without process cooperation. We have implemented the runtime system and applied it to three linear algebra algorithms: Cholesky, LU, and QR factorizations. Our experiments on both shared-memory machines (16, 32 cores) and distributed-memory machines (1024 cores) demonstrate that our runtime system is able to achieve good scalability. Furthermore, we provide analytical analysis to show why the tiled algorithms are scalable and the expected execution time.

This paper describes a novel approach to generate an optimized schedule to run threads on distributed shared memory (DSM) systems. The approach relies upon a binary instrumentation tool to automatically acquire the memory sharingrelationship between user-level threads by analyzing their memory trace. We introduce the concept of Affinity Graph to model the relationship. Expensive I/O for large trace files is completely eliminated by using an online graph creation scheme. We apply the technique of hierarchical graph partitioning and thread reordering to the affinity graph to determine an optimal thread schedule. We have performed experiments on an SGI Altix system. The experimental results show that our approach is able to reduce the totalexecution time by 10% to 38% for a variety of applications through the maximization of the data reuse within a single processor, minimization of the data sharing between processors, and a good load balance.

It is critical to provide high performance for scientific applications running on chip multi-processors (CMP). A CMP architecture often comprises a shared 12 cache and lower-level storages. The shared 12 cache can reduce the number of cache misses if the data are accessed in common by several threads, but it can also lead to performance degradation due to resource contention. Sometimes running threads on all cores can cause severe contention and increase the number of cache misses greatly. To investigate how the performance of a thread varies when running it concurrently with other threads on the remaining cores, we develop an analytical model to predict the number of misses on the shared L2 cache. In particular, we apply the model to thread-parallel numerical pro grams. We assume that all the threads compute homogeneous tasks and share a fully associative L2 cache. We use circular sequence profiling and stack processing techniques to analyze the L2 cache trace to predict the number of compulsory cache misses, capacity cache misses on shared data, and capacity cache misses on private data, respectively. Our method is able to predict the L2 cache performance for threads that have a global shared address space. For scientific applications, threads often have overlapping memory footprints. We use a cycle accurate simulator to validate the model with three scientific programs: dense matrix multiplication, blocked dense matrix multiplication, and sparse matrix-vector product. The average relative errors for the three experiments are 8.01%, 1.85%, and 2.41%, respectively.

Automatic pattern search in event traces is a powerful method to identify performance problems in parallel applications. We demonstrate that knowledge about the virtual topology, which defines logical adjacency relationships between processes, can be exploited to explain the occurrence of inefficiency patterns in terms of the parallelization strategy used in an application. We show correlations between higher-level events related to a parallel wavefront scheme and wait states identified by our pattern analysis. In addition, we visually expose relationships between pattern occurrences and the topological characteristics of the affected processes.

Performance tuning of parallel applications usually involves multiple experiments to compare the effects of different optimization strategies. This article describes an algebra that can be used to compare, integrate, and summarize performance data from multiple sources. The algebra consists of a data model to represent the data in a platform-independent fashion plus arithmetic operations to merge, subtract, and average the data from different experiments. A distinctive feature of this approach is its closure property, which allows processing and viewing all instances of the data model in the same way - regardless of whether they represent original or derived data - in addition to an arbitrary and easy composition of operations.