We consider optimal-scaling multigrid solvers for the linear systems that arise from the discretization of problems with evolutionary behavior. Typically, solution algorithms for evolution equations are based on a time-marching approach, solving sequentially for one time step after the other. Parallelism in these traditional time-integration techniques is limited to spatial parallelism. However, current trends in computer architectures are leading towards systems with more, but not faster, processors. Therefore, faster compute speeds must come from greater parallelism. One approach to achieve parallelism in time is with multigrid, but extending classical multigrid methods for elliptic operators to this setting is not straightforward. In this paper, we present a non-intrusive, optimal-scaling time-parallel method based on multigrid reduction (MGR). We demonstrate optimal-ity of our multigrid-reduction-in-time algorithm (MGRIT) for solving diffusion equations in two and three space dimensions in numerical experiments. Furthermore, through both parallel performance models and actual parallel numerical results, we show that we can achieve significant speedup in comparison to sequential time marching on modern architectures. 1. Introduction. One of the major challenges facing the computational science community with future architectures is that faster compute speeds must come from increased concurrency, since clock speeds are no longer increasing but core counts are going up sharply. As a consequence, traditional time marching is becoming a huge sequential bottleneck in time integration simulations in the following way: improving simulation accuracy by scaling up the spatial resolution requires a similar (or greater) increase in the temporal resolution, which is also required to maintain stability in explicit methods. As a result, numerical time integration involves many more time steps leading to long overall compute times, since parallelizing only in space limits concurrency. Solving for multiple time steps in parallel and, therefore, increasing concurrency would remove this time integration bottleneck. Because time is sequential in nature, the idea of simultaneously solving for multiple time steps is not intuitive. Yet it is possible, with work on this topic going back to as early as 1964 [33]. However, most research on this subject has been done within the past 30 years including [2, 7-10, 14-22, 25, 28, 31, 32, 38, 40-44]. One approach to achieve parallelism in time is with multigrid methods. The parareal in time method, introduced by Lions, Maday, and Turinici in [25], can be interpreted as a two-level multigrid method [16], even though the leading idea came from a spatial domain decomposition approach. The algorithm is optimal, but concurrency is limited since the coarse-grid solve is still sequential. Considering true multilevel (not two-level) schemes, only a few methods exhibit full multigrid optimality and concurrency such as [21, 42,43], and most are designed for specific problems or discretizations. Furt...

No abstract

Abstract. We consider optimal-scaling multigrid solvers for the linear systems that arise from the discretization of problems with evolutionary behavior. Typically, solution algorithms for evolution equations are based on a time-marching approach, solving sequentially for one time step after the other. Parallelism in these traditional time-integration techniques is limited to spatial parallelism. However, current trends in computer architectures are leading towards systems with more, but not faster, processors. Therefore, faster compute speeds must come from greater parallelism. One approach to achieve parallelism in time is with multigrid, but extending classical multigrid methods for elliptic operators to this setting is not straightforward. In this paper, we present a non-intrusive, optimal-scaling time-parallel method based on multigrid reduction (MGR). We demonstrate optimality of our multigrid-reduction-in-time algorithm (MGRIT) for solving diffusion equations in two and three space dimensions in numerical experiments. Furthermore, through both parallel performance models and actual parallel numerical results, we show that we can achieve significant speedup in comparison to sequential time marching on modern architectures.Key words. parabolic problems, reduction-based multigrid, multigrid-in-time, parareal AMS subject classifications. 65F10, 65M22, 65M551. Introduction. One of the major challenges facing the computational science community with future architectures is that faster compute speeds must come from increased concurrency, since clock speeds are no longer increasing but core counts are going up sharply. As a consequence, traditional time marching is becoming a huge sequential bottleneck in time integration simulations in the following way: improving simulation accuracy by scaling up the spatial resolution requires a similar (or greater) increase in the temporal resolution, which is also required to maintain stability in explicit methods. As a result, numerical time integration involves many more time steps leading to long overall compute times, since parallelizing only in space limits concurrency. Solving for multiple time steps in parallel and, therefore, increasing concurrency would remove this time integration bottleneck.Because time is sequential in nature, the idea of simultaneously solving for multiple time steps is not intuitive. Yet it is possible, with work on this topic going back to as early as 1964 [33]. However, most research on this subject has been done within the past 30 years including [2, 7-10, 14-22, 25, 28, 31, 32, 38, 40-44]. One approach to achieve parallelism in time is with multigrid methods. The parareal in time method, introduced by Lions, Maday, and Turinici in [25], can be interpreted as a two-level multigrid method [16], even though the leading idea came from a spatial domain decomposition approach. The algorithm is optimal, but concurrency is limited since the coarse-grid solve is still sequential. Considering true multilevel (not two-level) schemes, only a few method...

Algebraic multigrid methods solve sparse linear systems Ax = b by automatic construction of a multilevel hierarchy. This hierarchy is defined by grid transfer operators that must accurately capture algebraically smooth error relative to the relaxation method. We propose a methodology to improve grid transfers through energy minimization. The proposed strategy is applicable to Hermitian, non-Hermitian, definite, and indefinite problems. Each column of the grid transfer operator P is minimized in an energy-based norm while enforcing two types of constraints: a defined sparsity pattern and preservation of specified modes in the range of P. A Krylov-based strategy is used to minimize energy, which is equivalent to solving AP j = 0 for each column j of P , with the constraints ensuring a nontrivial solution. For the Hermitian positive definite case, a conjugate gradient (CG-)based method is utilized to construct grid transfers, while methods based on generalized minimum residual (GMRES) and CG on the normal equations (CGNR) are explored for the general case. The approach is flexible, allowing for arbitrary coarsenings, unrestricted sparsity patterns, straightforward long-distance interpolation, and general use of constraints, either user-defined or auto-generated. We conclude with numerical evidence in support of the proposed framework.

The need for parallelism in the time dimension is being driven by changes in computer architectures, where performance increases are now provided through greater concurrency, not faster clock speeds. This creates a bottleneck for sequential time marching schemes because they lack parallelism in the time dimension. Multigrid Reduction in Time (MGRIT) is an iterative procedure that allows for temporal parallelism by utilizing multigrid reduction techniques and a multilevel hierarchy of coarse time grids. MGRIT has been shown to be effective for linear problems, with speedups of up to 50 times. The goal of this work is the efficient solution of nonlinear problems with MGRIT, where efficiency is defined as achieving similar performance when compared to an equivalent linear problem. The benchmark nonlinear problem is the p-Laplacian, where p = 4 corresponds to a well-known nonlinear diffusion equation, and p = 2 corresponds to the standard linear diffusion operator, our benchmark linear problem. The key difficulty encountered is that the nonlinear time-step solver becomes progressively more expensive on coarser time levels, as the time-step size increases. To overcome such difficulties, multigrid research has historically targeted an accumulated body of experience regarding how to choose an appropriate solver for a specific problem type. To that end, this paper develops a library of MGRIT optimizations and modifications, most importantly an alternate initial guess for the nonlinear time-step solver and delayed spatial coarsening, that will allow many nonlinear parabolic problems to be solved with parallel scaling behavior comparable to the corresponding linear problem.