Engineering topics

Li, Xiaoye Sherry

Publications and source records attributed to Li, Xiaoye Sherry.

16 records

DOE OSTI2024

A graphics processing unit accelerated sparse direct solver and preconditioner with block low rank compression

We present the GPU implementation efforts and challenges of the sparse solver package STRUMPACK. The code is made publicly available on github with a permissive BSD license. STRUMPACK implements an approximate multifrontal solver, a sparse LU factorization which makes use of compression methods to accelerate time to solution and reduce memory usage. Multiple compression schemes based on rank-structured and hierarchical matrix approximations are supported, including hierarchically semi-separable, hierarchically off-diagonal butterfly, and block low rank. Here, in this paper, we present the GPU implementation of the block low rank (BLR) compression method within a multifrontal solver. Our GPU implementation relies on highly optimized vendor libraries such as cuBLAS and cuSOLVER for NVIDIA GPUs, rocBLAS and rocSOLVER for AMD GPUs and the Intel oneAPI Math Kernel Library (oneMKL) for Intel GPUs. Additionally, we rely on external open source libraries such as SLATE (Software for Linear Algebra Targeting Exascale), MAGMA (Matrix Algebra on GPU and Multi-core Architectures), and KBLAS (KAUST BLAS). SLATE is used as a GPU-capable ScaLAPACK replacement. From MAGMA we use variable sized batched dense linear algebra operations such as GEMM, TRSM and LU with partial pivoting. KBLAS provides efficient (batched) low rank matrix compression for NVIDIA GPUs using an adaptive randomized sampling scheme. The resulting sparse solver and preconditioner runs on NVIDIA, AMD and Intel GPUs. Interfaces are available from PETSc, Trilinos and MFEM, or the solver can be used directly in user code. We report results for a range of benchmark applications, using the Perlmutter system from NERSC, Frontier from ORNL, and Aurora from ALCF. For a high frequency wave equation on a regular mesh, using 32 Perlmutter compute nodes, the factorization phase of the exact GPU solver is about 6.5× faster compared to the CPU-only solver. The BLR-enabled GPU solver is about 13.8× faster than the CPU exact solver. For a collection of SuiteSparse matrices, the STRUMPACK exact factorization on a single GPU is on average 1.9× faster than NVIDIA’s cuDSS solver.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

ECP libraries and tools: An overview

The Exascale Computing Project (ECP) Software Technology and Co-Design teams addressed the growing complexities in high-performance computing (HPC) by developing scalable software libraries and tools that leverage exascale system capabilities. As we enter the exascale era, the need for reusable, optimized software solutions that can handle the unique challenges posed by these systems becomes increasingly important. The primary challenges the ECP teams faced were to create software libraries and tools that are performant on exascale architectures and portable and usable across diverse hardware platforms. Efforts addressed issues related to concurrent execution, memory management, and the integration of heterogeneous computing resources, such as GPUs from multiple vendors. The ECP’s strategy involved a structured development process encompassing the creation, optimization, and deployment of software in collaboration with industry, academia, and national laboratories. The project was organized into several technical areas: co-design of domain-specific suites with target applications, programming models and runtimes, development tools, mathematical libraries, data and visualization tools, and software ecosystem and delivery mechanisms. ECP has successfully developed a large portfolio of software libraries and tools that demonstrate significant improvements in performance and scalability on exascale systems. These products have been integrated into the Department of Energy’s computing facilities, supporting various scientific applications and ensuring robust performance across different hardware setups. ECP advancements in software development for exascale computing highlight the importance of a collaborative and adaptive approach to handling next-generation HPC systems complexities. The lessons learned emphasize the need for continuous engagement with end-users and vendors, and the importance of maintaining a balance between innovation and practical implementation. Future efforts will focus on ensuring scalability, keeping pace with rapid hardware advancements, and further enhancing the interoperability and usability of the software ecosystem. In conclusion, subsequent articles in this special issue provide in-depth discussions and case studies into specific library and tool efforts.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Unified Communication Optimization Strategies for Sparse Triangular Solver on CPU and GPU Clusters

This paper presents a unified communication optimization framework for sparse triangular solve (SpTRSV) algorithms on CPU and GPU clusters. The framework builds upon a 3D communication-avoiding (CA) layout of Px × Py × Pz processes that divides a sparse matrix into Pz submatrices, each handled by a Px × Py 2D grid with block-cyclic distribution. We propose three communication optimization strategies: First, a new 3D SpTRSV algorithm is developed, which trades the inter-grid communication and synchronization with replicated computation. This design requires only one inter-grid synchronization, and the inter-grid communication is efficiently implemented with sparse allreduce operations. Second, broadcast and reduction communication trees are used to reduce message latency of the intra-grid 2D communication on CPU clusters. Finally, we leverage GPU-initiated one-sided communication to implement the communication trees on GPU clusters. With these nested inter- and intra-grid communication optimization strategies, the proposed 3D SpTRSV algorithm can attain up to 3.45x speedups compared to the baseline 3D SpTRSV algorithm using up to 2048 Cori Haswell CPU cores. In addition, the proposed GPU 3D SpTRSV algorithm can achieve up to 6.5x speedups compared to the proposed CPU 3D SpTRSV algorithm with Pz up to 64. Finally it is remarkable that the proposed GPU 3D SpTRSV can scale to 256 GPUs using the Perlmutter system while the existing 2D SpTRSV algorithm can only scale up to 4 GPUs.

Sao, Piyush↗

DOE OSTI2023

Brief Announcement: Communication Optimal Sparse LU Factorization for Planar Matrices

We introduce a new parallel algorithm for solving sparse LU factorization of planar matrices, which commonly arise in the finite element method for 2D PDEs. Existing scalable methods, such as the multifrontal approach with subtree-to-subcube mapping by Gupta et al. [1] and right-looking with 3D mapping by Sao et al. [2] fail to achieve optimal communication costs for these matrices. Our new algorithm combines 3D mapping and subtree-to-subcube mapping to minimize communication costs while allowing trade-offs between extra memory and reduced communication. We demonstrate that our proposed algorithm attains the communication lower bound up to a factor of O(log log n) in the memory-optimal case and up to a factor of O(log P) in the memory-independent case for an n-dimensional planar sparse matrix on P processors.

Sao, Piyush↗

DOE OSTI2022

ECP Software Technology Capability Assessment Report V3.0

The Exascale Computing Project (ECP) Software Technology (ST) focus area is responsible for (1) developing critical software capabilities that will enable the successful execution of ECP applications and (2) providing key components of a productive and sustainable exascale computing ecosystem that will position the US Department of Energy (DOE) and the broader high-performance computing (HPC) community with a firm foundation for future extreme-scale computing capabilities. This ECP ST Capability Assessment Report (CAR) provides an overview and assessment of current ECP ST capabilities and activities, giving stakeholders and the broader HPC community information that can be used to assess ECP ST progress and plan their own efforts accordingly. ECP ST leaders commit to updating this document on regular basis (every 6–12 months). Highlights from this version of the report are presented here. This version of the CAR contains the following updates relative to the previous revision: (1) This report highlights the progress with the Extreme-scale Scientific Software Stack (E4S) efforts. In particular, this report discusses how E4S continues to gain traction as a first-class entity in the HPC ecosystem, enabling new conversations with users, facilities, vendors, other US agencies, and international partners. (2) The several-page summaries of each ECP Level 4 project were updated to reflect recent progress and next steps (Section 4). Of particular note are the experiences of our teams on early-access systems for Frontier. (3) The E4S is described further. E4S is now updated via quarterly releases. E4S is the primary integration and delivery vehicle for ECP ST capabilities (Section 2.1.1). (4) The ECP ST software development kit (SDK) effort further refined its groupings (Section 2.1.2). The ECP ST focus area represents the key bridge between exascale systems and the scientists developing applications that will run on those platforms. ECP ST efforts contribute to approximately 70 software products (Section 2.1.3) in six technical areas (Table 1). Since publishing the previous revision of the CAR, the team has continued to evolve the product dictionary of official product names, which enables more rigorous mapping of ECP ST deliverables to stakeholders (Section 2.1.4).

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

GSoFa: Scalable Sparse Symbolic LU Factorization on GPUs

Decomposing a matrix $\mathbf {A}$ into a lower matrix $\mathbf {L}$ and an upper matrix $\mathbf {U}$, which is also known as LU decomposition, is an essential operation in numerical linear algebra. For a sparse matrix, LU decomposition often introduces more nonzero entries in the $\mathbf {L}$ and $\mathbf {U}$ factors than in the original matrix. A symbolic factorization step is needed to identify the nonzero structures of $\mathbf {L}$ and $\mathbf {U}$ matrices. Attracted by the enormous potentials of the Graphics Processing Units (GPUs), an array of efforts have surged to deploy various LU factorization steps except for the symbolic factorization, to the best of our knowledge, on GPUs. This article introduces gSoFa, the first GPU-based symbolic factorization design with the following three optimizations to enable scalable LU symbolic factorization for nonsymmetric pattern sparse matrices on GPUs. First, here we introduce a novel fine-grained parallel symbolic factorization algorithm that is well suited for the Single Instruction Multiple Thread (SIMT) architecture of GPUs. Second, we tailor supernode detection into a SIMT friendly process and strive to balance the workload, minimize the communication and saturate the GPU computing resources during supernode detection. Third, we introduce a three-pronged optimization to reduce the excessive space consumption problem faced by multi-source concurrent symbolic factorization. Taken together, gSoFa achieves up to 31× speedup from 1 to 44 Summit nodes (6 to 264 GPUs) and outperforms the state-of-the-art CPU project, on average, by 5×. Notably, gSoFa also achieves up to 47 percent of the peak memory throughput of a V100 GPU in the Summit Supercomputer.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Resiliency in numerical algorithm design for extreme scale simulations

Here this work is based on the seminar titled ‘Resiliency in Numerical Algorithm Design for Extreme Scale Simulations’ held March 1–6, 2020, at Schloss Dagstuhl, that was attended by all the authors. Advanced supercomputing is characterized by very high computation speeds at the cost of involving an enormous amount of resources and costs. A typical large-scale computation running for 48 h on a system consuming 20 MW, as predicted for exascale systems, would consume a million kWh, corresponding to about 100k Euro in energy cost for executing 10 23 floating-point operations. It is clearly unacceptable to lose the whole computation if any of the several million parallel processes fails during the execution. Moreover, if a single operation suffers from a bit-flip error, should the whole computation be declared invalid? What about the notion of reproducibility itself: should this core paradigm of science be revised and refined for results that are obtained by large-scale simulation? Naive versions of conventional resilience techniques will not scale to the exascale regime: with a main memory footprint of tens of Petabytes, synchronously writing checkpoint data all the way to background storage at frequent intervals will create intolerable overheads in runtime and energy consumption. Forecasts show that the mean time between failures could be lower than the time to recover from such a checkpoint, so that large calculations at scale might not make any progress if robust alternatives are not investigated. More advanced resilience techniques must be devised. The key may lie in exploiting both advanced system features as well as specific application knowledge. Research will face two essential questions: (1) what are the reliability requirements for a particular computation and (2) how do we best design the algorithms and software to meet these requirements? While the analysis of use cases can help understand the particular reliability requirements, the construction of remedies is currently wide open. One avenue would be to refine and improve on system- or application-level checkpointing and rollback strategies in the case an error is detected. Developers might use fault notification interfaces and flexible runtime systems to respond to node failures in an application-dependent fashion. Novel numerical algorithms or more stochastic computational approaches may be required to meet accuracy requirements in the face of undetectable soft errors. These ideas constituted an essential topic of the seminar. The goal of this Dagstuhl Seminar was to bring together a diverse group of scientists with expertise in exascale computing to discuss novel ways to make applications resilient against detected and undetected faults. In particular, participants explored the role that algorithms and applications play in the holistic approach needed to tackle this challenge. This article gathers a broad range of perspectives on the role of algorithms, applications and systems in achieving resilience for extreme scale simulations. The ultimate goal is to spark novel ideas and encourage the development of concrete solutions for achieving such resilience holistically.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2021

Sparse Approximate Multifrontal Factorization with Butterfly Compression for High-Frequency Wave Equations

In this work, we present a fast and approximate multifrontal solver for large-scale sparse linear systems arising from finite-difference, finite-volume or finite-element discretization of high-frequency wave equations. The proposed solver leverages the butterfly algorithm and its hierarchical matrix extension for compressing and factorizing large frontal matrices via graph-distance guided entry evaluation or randomized matrix-vector multiplication-based schemes. Complexity analysis and numerical experiments demonstrate $\mathcal{O}(N\log^2 N)$ computation and $\mathcal{O}(N)$ memory complexity when applied to an $N\times N$ sparse system arising from 3D high-frequency Helmholtz and Maxwell problems.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Butterfly Factorization Via Randomized Matrix-Vector Multiplications

This paper presents an adaptive randomized algorithm for computing the butterfly factorization of an m × n matrix with m ≈ n provided that both the matrix and its transpose can be rapidly applied to arbitrary vectors. The resulting factorization is composed of O(log n) sparse factors, each containing O(n) nonzero entries. The factorization can be attained using O(n 3/2 log n) computation and O(n log n) memory resources. Furthermore, the proposed algorithm can be implemented in parallel and can apply to matrices with strong or weak admissibility conditions arising from surface integral equation solvers as well as multi-frontal-based finite-difference, finite-element, or finite-volume solvers. A distributed-memory parallel implementation of the algorithm demonstrates excellent scaling behavior.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

ECP Software Technology Capability Assessment Report

The Exascale Computing Project (ECP) Software Technology (ST) Focus Area is responsible for developing critical software capabilities that will enable successful execution of ECP applications, and for providing key components of a productive and sustainable Exascale computing ecosystem that will position the US Department of Energy (DOE) and the broader high performance (HPC) community with a firm foundation for future extreme-scale computing capabilities. This ECP ST Capability Assessment Report (CAR) provides an overview and assessment of current ECP ST capabilities and activities, giving stakeholders and the broader HPC community information that can be used to assess ECP ST progress and plan their own efforts accordingly. ECP ST leaders commit to updating this document on regular basis (every six to 12 months).

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

A parallel hierarchical blocked adaptive cross approximation algorithm

This article presents a low-rank decomposition algorithm based on subsampling of matrix entries. The proposed algorithm first computes rank-revealing decompositions of submatrices with a blocked adaptive cross approximation (BACA) algorithm, and then applies a hierarchical merge operation via truncated singular value decompositions (H-BACA). The proposed algorithm significantly improves the convergence of the baseline ACA algorithm and achieves reduced computational complexity compared to the traditional decompositions such as rank-revealing QR. Numerical results demonstrate the efficiency, accuracy, and parallel scalability of the proposed algorithm.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

A communication-avoiding 3D algorithm for sparse LU factorization on heterogeneous systems

We propose a new algorithm to improve the strong scalability of right-looking sparse LU factorization on distributed memory systems. Our 3D algorithm for sparse LU uses a three-dimensional MPI process grid, exploits elimination tree parallelism, and trades off increased memory for reduced per-process communication. We also analyze the asymptotic improvements for planar graphs (e.g., those arising from 2D grid or mesh discretizations) and certain non-planar graphs (specifically for 3D grids and meshes). For a planar graph with $n$ vertices, our algorithm reduces communication volume asymptotically in $n$ by a factor of $\mathscr{O}$ $\Big(\sqrt{log \ n}\Big)$ and latency by a factor of $\mathscr{O}$ $(log \ n)$. For nonplanar cases, our algorithm can reduce the per-process communication volume by 3× and latency by $\mathscr{O}$ $\Big(n^\frac13\Big)$ times. In all cases, the memory needed to achieve these gains is a constant factor. We implemented our algorithm by extending the 2D data structure used in SuperLU_DIST. Our new 3D code achieves empirical speedups up to 27× for planar graphs and up to 3.3× for non-planar graphs over the baseline 2D SuperLU_DIST when run on 24,000 cores of a Cray XC30. We extend the 3D algorithm for heterogeneous architectures by adding the Highly Asynchronous Lazy Offload (Halo) algorithm for co-processor offload. On 4096 nodes of a Cray XK7 with 32,768 CPU cores and 4096 Nvidia K20x GPUs, the 3D algorithm achieves empirical speedups up to 24× for planar graphs and 3.5× for non-planar graphs over the baseline 2D SuperLU_DIST with co-processor acceleration.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

A communication-avoiding 3D sparse triangular solver

We present a novel distributed memory algorithm to improve the strong scalability of the solution of a sparse triangular system. This operation appears in the solve phase of direct methods for solving general sparse linear systems, Ax = b. Our 3D sparse triangular solver employs several techniques, including a 3D MPI process grid, elimination tree parallelism, and data replication, all of which reduce the per-process communication when combined. We present analytical models to understand the communication cost of our algorithm and show that our 3D sparse triangular solver can reduce the per-process communication volume asymptotically by a factor of O(n1/4) and O(n1/6) for problems arising from the finite element discretizations of 2D "planar" and 3D "non-planar" PDEs, respectively. We implement our algorithm for use in SuperLU_DIST3D, using a hybrid MPI+OpenMP programming model. Our 3D triangular solve algorithm, when run on 12k cores of Cray XC30, outperforms the current state-of-the-art 2D algorithm by 7.2x for planar and 2.7x for the non-planar sparse matrices, respectively.

Sao, Piyush↗

DOE OSTI2018

A Study of Clustering Techniques and Hierarchical Matrix Formats for Kernel Ridge Regression

We present memory-efficient and scalable algorithms for kernel methods used in machine learning. Using hierarchical matrix approximations for the kernel matrix the memory requirements, the number of floating point operations, and the execution time are drastically reduced compared to standard dense linear algebra routines. We consider both the general H matrix hierarchical format as well as Hierarchically Semi-Separable (HSS) matrices. Furthermore, we investigate the impact of several preprocessing and clustering techniques on the hierarchical matrix compression. Effective clustering of the input leads to a ten-fold increase in efficiency of the compression. The algorithms are implemented using the STRUMPACK solver library. These results confirm that - with correct tuning of the hyperparameters - classification using kernel ridge regression with the compressed matrix does not lose prediction accuracy compared to the exact - not compressed - kernel matrix and that our approach can be extended to O(1M) datasets, for which computation with the full kernel matrix becomes prohibitively expensive. We present numerical experiments in a distributed memory environment up to 1,024 processors of the NERSC's Cori supercomputer using well-known datasets to the machine learning community that range from dimension 8 up to 784.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2018

A Communication-Avoiding 3D LU Factorization Algorithm for Sparse Matrices

We propose a new algorithm to improve the strong scalability of right-looking sparse LU factorization on distributed memory systems. Our 3D sparse LU algorithm uses a three-dimensional MPI process grid, aggressively exploits elimination tree parallelism and trades off increased memory for reduced per-process communication. We also analyze the asymptotic improvements for planar graphs (e.g., from 2D grid or mesh domains) and certain non-planar graphs (specifically for 3D grids and meshes). For planar graphs with n vertices, our algorithm reduces communication volume asymptotically in n by a factor of O{log n} and latency by a factor of O{log n}. For non-planar cases, our algorithm can reduce the per-process communication volume by 3× and latency by O{n^1/3} times. In all cases, the memory needed to achieve these gains is a constant factor. We implemented our algorithm by extending the 2D data structure used in superLU. Our new 3D code achieves speedups up to 27× for planar graphs and up to 3.3× for non-planar graphs over the baseline 2D superLU when run on 24,000 cores of a Cray XC30.

Sao, Piyush↗

DOE OSTI2017

Efficient Scalable Parallel Higher Order Direct MoM-SIE Method With Hierarchically Semiseparable Structures for 3-D Scattering

A novel fast scalable parallel algorithm is proposed for the solution of large 3-D scattering problems based on: 1) the double (geometrical and current-approximation) higher order (DHO) method of moments (MoM) in the surface integral equation (SIE) formulation and 2) a direct solver for dense linear systems utilizing hierarchically semiseparable (HSS) structures. Namely, an HSS matrix representation is used for compression, factorization, and solution of the system matrix. In addition, a rank-revealing QR decomposition for memory compression is used, with a stopping criterion in terms of the relative rank tolerance value. One method for geometrical preprocessing of the scatterers based on the cobblestone distance sorting technique is employed in order to enhance the HSS algorithm accuracy and parallelization. Numerical examples show how the accuracy of the DHO HSS-MoM-SIE method is easily controllable by using the relative tolerance for the matrix compression. Furthermore, the examples demonstrate low memory consumption, as well as much faster simulation time, when compared to the direct LU decomposition. The method enables dramatically faster monostatic scattering computations than iterative solvers and reduced number of unknowns when compared to low-order discretizations. Finally, great scalability of the algorithm is demonstrated on more than one thousand processes.

97 MATHEMATICS AND COMPUTING↗