Implementing Multi-GPU Scientific Computing Miniapps Across Performance Portable Frameworks

2024-11-27

2024 IEEE 42nd Central America and Panama Convention (CONCAPAN XLII)

AI Summary

To address performance portability challenges for scientific computing applications on heterogeneous architectures, this work systematically compares four portable programming models—Kokkos, OpenMP, RAJA, and OCCA—on a unified hardware platform (a single node with four A100 GPUs), evaluating end-to-end performance across two representative miniapps: N-body simulation and structured-grid computation. Our methodology unifies multi-GPU implementation across frameworks via just-in-time compilation, multi-GPU acceleration, distributed-memory parallelism, and efficient data synchronization. Results reveal that OCCA achieves superior performance for small-scale problems, while OpenMP exhibits significant bottlenecks on structured grids; substantial performance disparities exist among all frameworks. Critical bottlenecks are identified in inter-GPU communication overhead and reduction algorithm efficiency. This study establishes an empirical benchmark and methodological foundation for framework selection and co-optimization in heterogeneous high-performance scientific computing.

Technology Category

• Machine Learning: Hardware-aware ML
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Scientific computing in the exascale era demands increased computational power to solve complex problems across various domains. With the rise of heterogeneous computing architectures the need for vendor-agnostic, performance portability frameworks has been highlighted. Libraries like Kokkos have become essential for enabling high-performance computing applications to execute efficiently across different hardware platforms with minimal code changes. In this direction, this paper presents preliminary time-to-solution results for two representative scientific computing applications: an N-body simulation and a structured grid simulation. Both applications used a distributed memory approach and hardware acceleration through four performance portability frameworks: Kokkos, OpenMP, RAJA, and OCCA. Experiments conducted on a single node of the Polaris supercomputer using four NVIDIA A100 GPUs revealed significant performance variability among frameworks. OCCA demonstrated faster execution times for small-scale validation problems, likely due to JIT compilation, however its lack of optimized reduction algorithms may limit scalability for larger simulations while using its out of the box API. OpenMP performed poorly in the structured grid simulation most likely due to inefficiencies in inter-node data synchronization and communication. These findings highlight the need for further optimization to maximize each framework’s capabilities. Future work will focus on enhancing reduction algorithms, data communication, memory management, as wells as performing scalability studies, and a comprehensive statistical analysis to evaluate and compare framework performance.

Problem

• Evaluating performance portability frameworks for multi-GPU scientific computing applications
• Comparing execution efficiency across Kokkos, OpenMP, RAJA and OCCA frameworks
• Identifying optimization needs in reduction algorithms and data communication

Innovation

• Multi-GPU implementation across four portability frameworks
• Distributed memory approach with hardware acceleration
• Performance comparison using N-body and structured grid simulations