PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 29, 20260 citationsOpen Access

FLAMENCO: Performance Portability with OPS-DSL

View Full Paper
ALAshutosh Shankarrao LondheBTBen J R Thornber

Key Points

  • The research aims to re-engineer the FLAMENCO code for improved performance across various hardware systems using OPS.
  • Implemented parallelism through MPI-based domain decomposition.
  • Used OPS DSL for automatic parallel code generation and reduced complexity.
  • Maintained numerical fidelity while addressing data layout and stencil representation during porting.
  • Demonstrated efficient scaling to multi-billion-cell simulations across thousands of CPU cores.
  • Enabled diverse hardware targeting without requiring in-depth parallel programming expertise.

Abstract

FLAMENCO (FLow Analysis and Modeling ENvironment with COmbustion) is a well-established high-fidelity solver used for the simulation of turbulent reacting flows with detailed chemistry. The code solves the Navier–Stokes equations augmented with multi-species quasi-conservative advection equations using a structured finite-volume formulation. FLAMENCO employs fifth-order accurate spatial discretisation and a second-order extended stability Runge–Kutta time-integration scheme, enabling the resolution of both low- and high-Mach-number flow features, including shocks and turbulence. Parallelism is achieved through MPI-based domain decomposition, and the code has been demonstrated to scale efficiently to multi-billion-cell simulations on tens of thousands of CPU cores on leadership-class supercomputers. In this talk, we present ongoing research exploring the re-engineering of FLAMENCO using the OPS domain-specific language (DSL) with the aim of enabling performance portability across modern multicore CPUs and GPU-accelerated systems. OPS provides a high-level abstraction for structured- mesh applications, allowing the automatic generation of parallel code for multiple backends while significantly reducing the complexity of architecture-specific performance tuning. This approach enables domain scientists to target diverse hardware platforms without requiring detailed expertise in parallel programming models. We discuss the design considerations, challenges, and methodology involved in porting a large, mature Fortran-based combustion code such as FLAMENCO to OPS, including data layout, stencil representation, and preservation of numerical fidelity. This work lays the foundation for future performance and scalability studies of OPS-enabled FLAMENCO on heterogeneous HPC architectures and aims to facilitate efficient large-scale simulations of complex combustion phenomena such as flame–wall interactions, jets, and stratified reacting flows.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Londhe et al. (2026) studied this question.

synapsesocial.com/papers/69c8c35cde0f0f753b39e247https://doi.org/10.5281/zenodo.19257114
Ask AI
Helpful
Bookmark
Share
View Full Paper