-
A Numerical Investigation of Matrix-Free Implicit Time-Step** Methods for Large CFD Simulations
Authors:
Arash Sarshar,
Paul Tranquilli,
Brent Pickering,
Andrew McCall,
Adrian Sandu,
Christopher J. Roy
Abstract:
This paper is concerned with the development and testing of advanced time-step** methods suited for the integration of time-accurate, real-world applications of computational fluid dynamics (CFD). The performance of several time discretization methods is studied numerically with regards to computational efficiency, order of accuracy, and stability, as well as the ability to treat effectively sti…
▽ More
This paper is concerned with the development and testing of advanced time-step** methods suited for the integration of time-accurate, real-world applications of computational fluid dynamics (CFD). The performance of several time discretization methods is studied numerically with regards to computational efficiency, order of accuracy, and stability, as well as the ability to treat effectively stiff problems. We consider matrix-free implementations, a popular approach for time-step** methods applied to large CFD applications due to its adherence to scalable matrix-vector operations and a small memory footprint. We compare explicit methods with matrix-free implementations of implicit, linearly-implicit, as well as Rosenbrock-Krylov methods. We show that Rosenbrock-Krylov methods are competitive with existing techniques excelling for a number of problem types and settings.
△ Less
Submitted 30 September, 2017; v1 submitted 22 July, 2016;
originally announced July 2016.
-
Application of approximate matrix factorization to high order linearly implicit Runge-Kutta methods
Authors:
Hong Zhang,
Adrian Sandu,
Paul Tranquilli
Abstract:
Linearly implicit Runge-Kutta methods with approximate matrix factorization can solve efficiently large systems of differential equations that have a stiff linear part, e.g. reaction-diffusion systems. However, the use of approximate factorization usually leads to loss of accuracy, which makes it attractive only for low order time integration schemes. This paper discusses the application of approx…
▽ More
Linearly implicit Runge-Kutta methods with approximate matrix factorization can solve efficiently large systems of differential equations that have a stiff linear part, e.g. reaction-diffusion systems. However, the use of approximate factorization usually leads to loss of accuracy, which makes it attractive only for low order time integration schemes. This paper discusses the application of approximate matrix factorization with high order methods; an inexpensive correction procedure applied to each stage allows to retain the high order of the underlying linearly implicit Runge-Kutta scheme. The accuracy and stability of the methods are studied. Numerical experiments on reaction-diffusion type problems of different sizes and with different degrees of stiffness illustrate the efficiency of the proposed approach.
△ Less
Submitted 17 August, 2014; v1 submitted 15 August, 2014;
originally announced August 2014.
-
Indexed Labels for Loop Iteration Dependent Costs
Authors:
Paolo Tranquilli
Abstract:
We present an extension to the labelling approach, a technique for lifting resource consumption information from compiled to source code. This approach, which is at the core of the annotating compiler from a large fragment of C to 8051 assembly of the CerCo project, looses preciseness when differences arise as to the cost of the same portion of code, whether due to code transformation such as loop…
▽ More
We present an extension to the labelling approach, a technique for lifting resource consumption information from compiled to source code. This approach, which is at the core of the annotating compiler from a large fragment of C to 8051 assembly of the CerCo project, looses preciseness when differences arise as to the cost of the same portion of code, whether due to code transformation such as loop optimisations or advanced architecture features (e.g. cache). We propose to address this weakness by formally indexing cost labels with the iterations of the containing loops they occur in. These indexes can be transformed during the compilation, and when lifted back to source code they produce dependent costs.
The proposed changes have been implemented in CerCo's untrusted prototype compiler from a large fragment of C to 8051 assembly.
△ Less
Submitted 11 June, 2013;
originally announced June 2013.
-
Standardization in resource lambda-calculus
Authors:
Maurizio Dominici,
Simona Ronchi Della Rocca,
Paolo Tranquilli
Abstract:
The resource calculus is an extension of the lambda-calculus allowing to model resource consumption. It is intrinsically non-deterministic and has two general notions of reduction - one parallel, preserving all the possible results as a formal sum, and one non-deterministic, performing an exclusive choice at every step. We prove that the non-deterministic reduction enjoys a notion of standardizati…
▽ More
The resource calculus is an extension of the lambda-calculus allowing to model resource consumption. It is intrinsically non-deterministic and has two general notions of reduction - one parallel, preserving all the possible results as a formal sum, and one non-deterministic, performing an exclusive choice at every step. We prove that the non-deterministic reduction enjoys a notion of standardization, which is the natural extension with respect to the similar one in classical lambda-calculus. The full parallel reduction only enjoys a weaker notion of standardization instead. The result allows an operational characterization of may-solvability, which has been introduced and already characterized (from the syntactical and logical points of view) by Pagani and Ronchi Della Rocca.
△ Less
Submitted 17 November, 2012;
originally announced November 2012.
