pith. sign in

arxiv: 1807.03032 · v3 · pith:3CBKVU5Tnew · submitted 2018-07-09 · 💻 cs.MS

Architecture and performance of Devito, a system for automated stencil computation

Fabio Luporini , Michael Lange , Mathias Louboutin , Navjot Kukreja , Jan H\"uckelheim , Charles Yount , Philipp Witte , Paul H. J. Kelly
show 2 more authors
Felix J. Herrmann Gerard J. Gorman
This is my paper
classification 💻 cs.MS
keywords devitoperformancestencilcompilerequationsoptimizationsapplicationsarchitecture
0
0 comments X
read the original abstract

Stencil computations are a key part of many high-performance computing applications, such as image processing, convolutional neural networks, and finite-difference solvers for partial differential equations. Devito is a framework capable of generating highly-optimized code given symbolic equations expressed in Python, specialized in, but not limited to, affine (stencil) codes. The lowering process---from mathematical equations down to C++ code---is performed by the Devito compiler through a series of intermediate representations. Several performance optimizations are introduced, including advanced common sub-expressions elimination, tiling and parallelization. Some of these are obtained through well-established stencil optimizers, integrated in the back-end of the Devito compiler. The architecture of the Devito compiler, as well as the performance optimizations that are applied when generating code, are presented. The effectiveness of such performance optimizations is demonstrated using operators drawn from seismic imaging applications.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Joint elastic full waveform inversion of multi-component geophone and distributed acoustic sensing data

    physics.geo-ph 2026-07 unverdicted novelty 6.0

    A VSS-based joint FWI framework enables direct multi-deployment inversion of geophone and DAS data, yielding more accurate elastic parameter recovery than single-sensor cases on Marmousi benchmarks when sensors provid...

  2. Stencil Computations on Cerebras Wafer-Scale Engine

    cs.DC 2026-05 unverdicted novelty 6.0

    CStencil on the WSE-3 achieves up to 342x speedup for 2D stencils versus an adapted single-precision GPU solver and saturates both compute and on-chip memory bandwidth.

  3. Automatic Differentiation for Adjoint Stencil Loops

    cs.DC 2019-07 unverdicted novelty 6.0

    A method for adjoint differentiation of stencil loops that preserves their structure and parallelizability via combined AD and loop transformations, released as the PerforAD tool with seismic and CFD test cases.

  4. Arbitrary high order shaped stencils for time domain finite difference schemes in seismic wave propagation

    math.NA 2026-06 unverdicted novelty 4.0

    A general mathematical framework is proposed for deriving arbitrary high-order shaped stencils in finite difference schemes for acoustic wave propagation, with tests indicating benefits for dispersion-optimized cross ...

  5. Investigating the OPS intermediate representation to target GPUs in the Devito DSL

    cs.MS 2019-06 unverdicted novelty 3.0

    Integration of OPS intermediate representation as a GPU backend in Devito DSL yields speedups over the core backend for structured-mesh finite-difference applications.