Trilinos is a community-developed, open-source software framework that facilitates building large-scale, complex, multiscale, multiphysics simulation code bases for scientific and engineering problems. Since the Trilinos framework has undergone substantial changes to support new applications and new hardware architectures, this document is an update to “An Overview of the Trilinos project” by Heroux et al. (ACM Transactions on Mathematical Software, 31(3):397–423, 2005). It describes the design of Trilinos, introduces its new organization in product areas, and highlights established and new features available in Trilinos. Particular focus is put on the modernized software stack based on the Kokkos ecosystem to deliver performance portability across heterogeneous hardware architectures. This article also outlines the organization of the Trilinos community and the contribution model to help onboard interested users and contributors.

We present a fault model designed to bring out the "worst" in iterative solvers based on mathematical properties. Our model introduces substantially higher overhead, but smaller variance, than a fault model based on random bit flips. We also relate the statistics from our experiments back to the solvers' configuration, and briefly address the computational effort that each model requires. Our approach requires significantly fewer resources, while punishing our solvers with undetectable errors that require notable overhead for recovery. This work also illustrates the robustness of our resilient algorithms: Not only do we make forward progress in the presence of pathological faults, we always obtain the correct answer.

Increasing parallelism and transistor density, along with increasingly tighter energy and peak power constraints, may force exposure of occasionally incorrect computation or storage to application codes. Silent data corruption (SDC) will likely be infrequent, yet one SDC suffices to make numerical algorithms like iterative linear solvers cease progress towards the correct answer. Thus, we focus on resilience of the iterative linear solver GMRES to a single transient SDC. We derive inexpensive checks to detect the effects of an SDC in GMRES that work for a more general SDC model than presuming a bit flip. Our experiments show that when GMRES is used as the inner solver of an inner-outer iteration, it can "run through" SDC of almost any magnitude in the computationally intensive orthogonalization phase. That is, it gets the right answer using faulty data without any required roll back. Those SDCs which it cannot run through, get caught by our detection scheme.

The Generalized Minimum Residual (GMRES) method is one of the most widely-used iterative methods for solving nonsymmetric linear systems of equations. In recent years, techniques to avoid communication in GMRES have gained attention because in comparison to floating-point operations, communication is becoming increasingly expensive on modern computers. Since graphics processing units (GPUs) are now becoming crucial component in computing, we investigate the effectiveness of these techniques on multicore CPUs with multiple GPUs. While we present the detailed performance studies of a matrix powers kernel on multiple GPUs, we particularly focus on orthogonalization strategies that have a great impact on both the numerical stability and performance of GMRES, especially as the matrix becomes sparser or ill-conditioned. We present the experimental results on two eight-core Intel Sandy Bridge CPUs with three NDIVIA Fermi GPUs and demonstrate that significant speedups can be obtained by avoiding communication, either on a GPU or between the GPUs. As part of our study, we investigate several optimization techniques for the GPU kernels that can also be used in other iterative solvers besides GMRES. Hence, our studies not only emphasize the importance of avoiding communication on GPUs, but they also provide insight about the effects of these optimization techniques on the performance of the sparse solvers, and may have greater impact beyond GMRES.

Krylov subspace projection methods are widely used iterative methods for solving large-scale linear systems of equations. Researchers have demonstrated that communication avoiding (CA) techniques can improve Krylov methods' performance on modern computers, where communication is becoming increasingly expensive compared to arithmetic operations. In this paper, we extend these studies by two major contributions. First, we present our implementation of a CA variant of the Generalized Minimum Residual (GMRES) method, called CAGMRES, for solving no symmetric linear systems of equations on a hybrid CPU/GPU cluster. Our performance results on up to 120 GPUs show that CA-GMRES gives a speedup of up to 2.5x in total solution time over standard GMRES on a hybrid cluster with twelve Intel Xeon CPUs and three Nvidia Fermi GPUs on each node. We then outline a domain decomposition framework to introduce a family of preconditioners that are suitable for CA Krylov methods. Our preconditioners do not incur any additional communication and allow the easy reuse of existing algorithms and software for the sub domain solves. Experimental results on the hybrid CPU/GPU cluster demonstrate that CA-GMRES with preconditioning achieve a speedup of up to 7.4x over CAGMRES without preconditioning, and speedup of up to 1.7x over GMRES with preconditioning in total solution time. These results confirm the potential of our framework to develop a practical and effective preconditioned CA Krylov method.

Orthogonalization consumes much of the run time of many iterative methods for solving sparse linear systems and eigenvalue problems. Commonly used algorithms, such as variants of Gram-Schmidt or Householder QR, have performance dominated by communication. Here, "communication" includes both data movement between the CPU and memory, and messages between processors in parallel. Our Tall Skinny QR (TSQR) family of algorithms requires asymptotically fewer messages between processors and data movement between CPU and memory than typical orthogonalization methods, yet achieves the same accuracy as Householder QR factorization. Furthermore, in block orthogonalizations, TSQR is faster and more accurate than existing approaches for orthogonalizing the vectors within each block ("normalization"). TSQR's rank-revealing capability also makes it useful for detecting deflation in block iterative methods, for which existing approaches sacrifice performance, accuracy, or both. We have implemented a version of TSQR that exploits both distributed-memory and shared-memory parallelism, and supports real and complex arithmetic. Our implementation is optimized for the case of orthogonalizing a small number (5 -- 20) of very long vectors. The shared-memory parallel component uses Intel's Threading Building Blocks, though its modular design supports other shared-memory programming models as well, including computation on the GPU. Our implementation achieves speedups of 2 times or more over competing orthogonalizations. It is available now in the development branch of the Trilinos software package, and will be included in the 10.8 release.

Data communication within the memory system of a single processor node and between multiple nodes in a system is the bottleneck in many iterative sparse matrix solvers like CG and GMRES. Here k iterations of a conventional implementation perform k sparse-matrix-vector-multiplications and Ω(k) vector operations like dot products, resulting in communication that grows by a factor of Ω(k) in both the memory and network. By reorganizing the sparse-matrix kernel to compute a set of matrix-vector products at once and reorganizing the rest of the algorithm accordingly, we can perform k iterations by sending O(log P) messages instead of O(k · log P) messages on a parallel machine, and reading the matrix A from DRAM to cache just once, instead of k times on a sequential machine. This reduces communication to the minimum possible. We combine these techniques to form a new variant of GMRES. Our shared-memory implementation on an 8-core Intel Clovertown gets speedups of up to 4.3x over standard GMRES, without sacrificing convergence rate or numerical stability.

The performance of sparse iterative solvers is typically limited by sparse matrix-vector multiplication, which is itself limited by memory system and network performance. As the gap between computation and communication speed continues to widen, these traditional sparse methods will suffer. In this paper we focus on an alternative building block for sparse iterative solvers, the "matrix powers kernel" [x, Ax, A2x, ..., Akx], and show that by organizing computations around this kernel, we can achieve near-minimal communication costs. We consider communication very broadly as both network communication in parallel code and memory hierarchy access in sequential code. In particular, we introduce a parallel algorithm for which the number of messages (total latency cost) is independent of the power k, and a sequential algorithm, that reduces both the number and volume of accesses, so that it is independent of k in both latency and bandwidth costs. This is part of a larger project to develop "communication-avoiding Krylov subspace methods," which also addresses the numerical issues associated with these methods. Our algorithms work for general sparse matrices that "partition well". We introduce parallel performance models of matrices arising from 2D and 3D problems and show predicted speedups over a conventional algorithm of up to 7times on a petaflop-scale machine and up to 22times on computation across the grid. Analogous sequential performance models of the same problems predict speedups over a conventional algorithm of up to 10times on an out-of-core implementation, and up to 2.5times when we use our ideas to reduce off-chip latency and bandwidth to DRAM. Finally, we validate the model on an out-of-core sequential implementation and measured a speedup of over 3times, which is close to the predicted speedup.